The Political Consequences of Ethnically Targeted Incarceration: Evidence from Japanese-American Internment During WWII Faculty Research Working Paper Series
Mayya Komisarchik University of Rochester Maya Sen Harvard Kennedy School Yamil R. Velez Columbia University
July 2020 RWP20-021
Visit the HKS Faculty Research Working Paper Series at: https://www.hks.harvard.edu/research-insights/publications?f%5B0%5D=publication_types%3A121 The views expressed in the HKS Faculty Research Working Paper Series are those of the author(s) and do not necessarily reflect those of the John F. Kennedy School of Government or of Harvard University. Faculty Research Working Papers have not undergone formal review and approval. Such papers are included in this series to elicit feedback and to encourage debate on important public policy challenges. Copyright belongs to the author(s). Papers may be downloaded for personal use only.
www.hks.harvard.edu e Political Consequences of Ethnically Targeted Incarceration: Evidence from Japanese-American Internment During WWII∗
Mayya Komisarchik† Maya Sen‡ Yamil R. Velez§
March 5, 2020
Abstract What are the downstream political consequences of state activity explicitly targeting a minority group? is question is well studied in the comparative context, but less is known about the eects of explicitly racist state activity in liberal democracies such as the United States. We investigate this question by looking at an important event in American history— the mass internment of people of Japanese ancestry during World War II. We nd that Japanese Americans who were interned or had family who were interned are signicantly less politi- cally engaged and that this demobilizing eect increases with internment length. We also nd that camp experience maers: those who went to camps that witnessed violence or strikes experienced sharper declines. Taken together, our ndings contribute to a growing literature documenting the demobilizing eects of ethnically targeted incarceration and expand our understanding of these forces within the U.S.
Word Count: 11,419
∗We are grateful to Michael Hankinson, Alexander Hertel-Fernandez, Boris Heersink, Katherine Krimmel, Mahew Lacombe, Sarah Merchant, and Michael Miller for helpful feedback. Special thanks to Danny Shoag for conversations on an earlier iteration. Comments and suggestions welcome. †Department of Political Science, University of Rochester. ‡John F. Kennedy School of Government, Harvard University. §Department of Political Science, Columbia University. 1 Introduction
As immigrant populations have expanded across liberal democracies, there has been a con- current rise in the use of punitive policies targeting immigrant and ethnic minority populations. Among various methods, scholars have identied putatively illiberal practices such as inde- nite detention (Welch, 2002), corralling of unauthorized immigrants into crowded holding facili- ties (Bosworth and Turnbull, 2014), and reinforcement of border fencing and barriers that create hostile conditions for migrants (Roche, 2014). ese contemporary policies highlight important tensions within liberal democracies over whether the rights and protections conferred to long- standing majority groups extend to immigrants and their progeny; these tensions have revealed themselves throughout history and continue to raise questions about the universal scope of demo- cratic principles (Epstein, 2016). e internment by the U.S. government of people of Japanese descent is a key historical case that frames the dierence between American democratic principles and practice. In June of 1942, approximately 120,000 people of Japanese descent, the majority of whom were U.S. citizens living along the U.S. West Coast, were sent to internment camps throughout the American interior (Weglyn, 1976). Although this internment faced legal challenges as well as civil disobedience, it was not until the end of World War II that internment camps were shut down. By then, hundreds of thousands of people had been displaced and their lives severely disrupted by the state. is included not just adults, but also teenagers and children—many of whom spent formative years living in military camps. Despite both the historical signicance of this event and recent growing use of detention centers along the U.S.-Mexico border and elsewhere in the world (Sampson and Mitchell, 2013), relatively lile research in political science has examined the downstream political consequences of such large-scale racial targeting within liberal democracies. Many studies in comparative pol- itics have examined forced migration and the internment of racial or ethnic minorities, coming to the conclusion that such migrations serve to further iname inter-ethnic conict and reduce the trust that minority groups express toward not just majority groups, but also toward the state
1 (e.g., Lupu and Peisakhin, 2017; Zhukov and Talibova, 2018). In addition, a growing literature within American politics has examined the crippling impact of incarceration and other punitive encounters with governments on aected populations (e.g., Lerman and Weaver, 2014; White, 2019b). is literature—focused primarily on penal institutions (as opposed to military ones)— has found that state-sponsored incarceration depresses the political engagement not just of those incarcerated but also that of their extended families (White, 2019a). e context of Japanese internment provides an important instance within American history to explore the eects of a systematic and ethnically targeted policy on subsequent political ai- tudes and behavior. First, the internment of Japanese Americans during WWII was a large-scale government activity, puing it on par with ethnically driven state activity in a comparative con- text. In addition, such episodes are rarer in liberal democracies, thereby aording an opportunity to assess the impact of explicit ethnic targeting outside the more well-studied context of author- itarian or weak states. Moreover, the legacy of Japanese-American interment provides useful variation: not only was there variation in who was interned, but, conditional on original location, Japanese-American families were mostly exogenously assigned to one of 10 War Relocation Au- thority (WRA) camps throughout the U.S. We leverage this variation to explore how conditions associated with internment impacted Japanese-Americans’ subsequent aitudes and behavior. We nd several factors suggesting that this racist state-sanctioned activity has had extensive downstream repercussions. First, exploring which individuals did or did not live within the ex- clusion zone, we nd that being interned or having family members who were interned has had a lasting, large, and signicant depressing eect on political engagement—one that cannot be ex- plained by factors that plausibly covary with internment status (such as lack of military service or dierential income). Second, conditional on internment, we nd that an additional year of internment is associated with a sizable decrease in political trust and engagement. ird, to ex- plore the mechanism by which internment could be demobilizing, we examine the nature of the camps themselves. We nd that being assigned to a camp in which violence or unrest erupted re- sulted in greater political disengagement among those interned and their families. is suggests
2 to us that a key pathway to disengagement is the condition of the camp itself—with more unrest among internees being linked to lower engagement. Surprisingly, we do not nd similar eects for exposure to militaristic environments or the cultural or political environment in camps’ sur- rounding areas. is suggests that the collective negative social experience of internees, rather than exposure to a particular environment, forms the foundation of the demobilizing eect. is paper makes several contributions. First, we link disparate literatures from comparative politics with scholarship on American politics, explaining how state-sponsored racial targeting— even within a large liberal democracy—can have negative downstream consequences on political engagement among those directly aected and also their families. Second, our research shows that the severity of these experiences maers, such that more adverse collective conditions in sites of state captivity produce more sizable eects over time. ird, our study provides an unprecedented opportunity to assess theories of “carceral contact” using a case where the psychological linkages between the internment and the government’s role are strong. Our ndings therefore have strong implications for governments’ current-day use of detention centers, including those conning migrants. Lastly, our ndings engage a growing literature on Japanese-American public opinion, questioning and complicating the link between Japanese-Americans’ historically high political engagement and the history of internment. is paper proceeds as follows. We rst provide a review of the comparative literature on ethnic conict, in tandem with a growing literature on the political consequences of the American “carceral state” as well as the literature on Japanese-American political behavior. We then provide context on the WWII Japanese-American internment process and provide an overview of our data, which include novel data on the conditions of the camps and their surroundings. We next present our main results showing that direct and family exposure to the internment process predicts subsequent political disengagement and that, among those interned, the length of the internment does as well. ese results are not explained by alternative explanations, such as dierences in terms of military service or economic success. We next provide an analysis showing that social unrest appears to play a key role in furthering this demobilization. Finally, we demonstrate that
3 our design assumptions are robust to several challenges as well as alternative explanations, such as the possibility that ndings are being driven by exposure to surrounding camp environments. We conclude by noting how our work informs other ndings on carceral policies in western democracies.
2 How Ethnically Targeted State Activity Could Impact Minority Group Behavior
States have historically exercised control over ethnic minority groups via targeted repression and violence during periods of social upheaval, mass migration, or war (Levy, 1988; Arendt, 1973; Reeves, 2015). Scholars have explored the political and psychological eects of these eorts, with some studies revealing that repression can politically mobilize targeted minority groups (LeBas, 2006; Davenport, 2005; Blaman, 2009; Bellows and Miguel, 2009) and others suggesting demobi- lizing eects (Lyall, 2009). While this literature has shed light on the potential political impacts of violence and repression on targeted communities, recent work has drawn aention to the impor- tance of regime strength and whether repression eorts are carried out by state actors. As Zhukov and Talibova (2018) note, empirical examinations of the repression-mobilization link have tradi- tionally selected cases of ethnic targeting perpetrated by non-state actors or weak states. However, we know surprisingly lile about the impact of large-scale repression eorts by “strong” states on minority groups’ political participation. Focusing on the mass deportation of ethnic minorities during the Stalin Era, Lupu and Peisakhin (2017) and Rozenas, Schue and Zhukov (2017) nd that punitive encounters with the Soviet Union during the 20th century con- tinue to have positive and durable eects on group aachments and support for minority leaders among targeted groups.1 Moreover, these studies nd that exposure to forced migration durably
1Among the various methods states use to repress ethnic groups, their capacity to move groups is especially ef- fective. As McGarry (1998) notes, forced migration can dilute the power of ethnic minorities, allow states to reassert sovereignty over disputed territories, increase protection against external threats, and reduce ethnic conict. Unsur- prisingly, then, studies exploring the eects of state-directed ethnic targeting have gravitated toward forced migration as a relevant opportunity to test theories of repression.
4 aects voting paerns, such that those who experienced violence in the past report lower levels of support for pro-Russian parties in the contemporary period.
2.1 Ethnic Targeting in Liberal Democracies
A potential limitation of this literature has been its focus on authoritarian regimes—where opportunities to express grievances are limited and state repercussion can be severe. Even within liberal democracies, however, minority groups have collided with punitive institutions and experi- enced the consequences of ethnic targeting (Zakaria, 1997; Wahab Twibell, 2004; Mann, 2005; Soss and Schram, 2007; Fouka, 2019).2 For example, an arena that possibly speaks to our inquiry is the contemporary status of Jews in Europe, who were targets of ethnic atrocities by fascist Germany but many of whom continue to live in Europe today. Although some recent work has found eects of concentration camps on local non-Jewish populations (Homola, Pereira and Tavits, 2020), the literature on the downstream eects of the Holocaust on Jewish political participation is scarce; what limited scholarship there is suggests that Jews may be more politically engaged than the general population (e.g., Schnapper, Bordes-Benayoun and Raphael, 2011). Moving to the present day, the past few decades have seen liberal democracies banning forms of religious dress (Nanwani, 2011), increasing the surveillance and deportation of migrants (Kalhan, 2013), and sidestepping due process in the name of national security (Radack, 2004). e detention of migrants in camps has become more widespread across Australia, the U.S., and Europe (Samp- son and Mitchell, 2013). Despite increases in such policies targeting particular ethnic groups, it is unclear whether the eects of ethnic targeting and repression translate across dierent kinds of political systems. On the one hand, “threat-mobilization” studies in the U.S. have shown that policies targeting ethnic groups can increase minority political participation (Pantoja, Ramirez and Segura, 2001; Bowler, Nicholson and Segura, 2006; White, 2016). Consistent with some of the existing literature on ethnic violence (Lupu and Peisakhin, 2017), prominent theories posit that
2Fouka (2019), for instance, studies the impact of discrimination on assimilation paerns among immigrants, and nds that German immigrants during World War I were more likely to increase their “assimilation investments” relative to other immigrant groups.
5 punitive policies increase the salience of ethnic identities, thereby strengthening the link between group aachments and voting paerns (White, 2016).3 On the other hand, research on “custodial citizenship”—which includes literature exploring the downstream consequences of state-sponsored incarceration on minority populations—has found that direct experience with racially disparate criminal justice policies can have demobi- lizing eects (Uggen and Manza, 2002; Hjalmarsson and Lopez, 2010; Burch, 2013; Lee, Porter and Comfort, 2014; Epp, Maynard-Moody and Haider-Markel, 2014; Walker, 2014; Kupchik and Catlaw, 2015). According to this perspective, contact with the criminal justice system serves as a political socialization experience that erodes trust in government by exposing minority groups to aggressive elements of otherwise democratic systems (Lerman and Weaver, 2014). We note that, in contrast to cases of ethnic targeting in comparative contexts (where state- directed eorts are intended to encompass entire groups), this literature in American politics has focused on policies that target subsets of minority populations.4 However, this complicates our ability to assess the political eects of punitive policies because “policy recipients” oen dier in important ways from other group members who do not directly experience the policy (Maltby, 2017). e potential mismatch between contemporary cases of racial or ethnic targeting in liberal democracies and similar cases in comparative politics muddles the transportability of theories and expectations across political systems. In addition, existing studies assess contact with punitive institutions in a binary fashion (i.e., contact v. no contact). However, as Weaver, Hacker and Wildeman (2014, 19) note, scholars have yet to “move beyond treating incarceration as a uniform treatment” and leverage “variation in the character of custodial interactions, not just their incidence.” We take up this invitation in our analyses below.
3See Dehdari and Gehring (2018) for a similar argument in the context of regional identities. 4For instance, despite signicant racial disparities in the criminal justice system (Soss and Weaver, 2017), a mi- nority of blacks and Latinos experience incarceration. Moreover, restrictionist immigration policies tend to target unauthorized immigrants who comprise a relatively small proportion of the broader immigrant population.
6 2.2 Japanese-Americans, Internment, and Expectations
e case of Japanese-American internment during WWII makes an urgent and compelling case to study these topics. First, unlike instances in authoritarian regimes, the Japanese-American case took place in a relatively well-functioning, modern, liberal democracy.5 Second, the intern- ment of Japanese Americans involved the explicit targeting of an entire ethnic group, a contrast with ostensibly race-neutral targeting (such as policing or even immigrant detention centers). Lastly, Japanese Americans remained in the U.S. following their internment, enabling us to eval- uate subsequent political participation. In this regard, Japanese-American political behavior has been the subject of robust scholarly research, with several studies nding a high baseline level of political engagement. For exam- ple, Wong et al. (2011, p. 18-20), report that, among Asian Americans, “Japanese Americans are the likeliest group to be registered to vote and to report voting,” with registration, turnout, and engagement rates far outpacing U.S. averages (Wong et al. (2011), Table 1.1). is activism has translated into political oceholding: “many Japanese Americans ran for elected oce and, by 2008, the majority of Asian American elected ocials were of Japanese descent” (Wong et al., 2011, p. 48). However, the same authors also report that “among national-origin groups…Japanese (11 percent) are the most likely to report being a victim of a hate crime” (Wong et al., 2011, p. 169). Several scholars have linked political engagement among Japanese Americans back to intern- ment. For example,Wong et al. (2011) argue that Japanese Americans as a group “are characterized by relatively high socioeconomic status, as well as stark historical experiences of racial discrim- ination, including the internment of Japanese Americans during World War II” (p. 180). Within cultural anthropology, Takezawa (1991) points to the movement for redress as a unifying event for many third-generation Japanese Americans; “the ethnicity of the Sansei today is constructed not merely from racial and cultural markers of pre-war days,” she notes, “but from a sense of
5Scholars have noted subnational authoritarianism in the U.S. South during the early 20th Century (e.g., Mickey, 2015), not to mention the proceeding 250 years of chael slavery. In addition, American history is replete with other examples of racial targeting, including the targeting of Native Americans, Asian immigrants, and Latino/as, as well as more localized targeting of certain white immigrants. Religious minorities (for example, Mormons) also have histori- cally been targeted by the state.
7 suering of their forebears who experienced internment” (p. 41). As one of the Japanese Ameri- cans that she interviews notes, “My parents went to the camp, that aected me indirectly. Now I have become closer to the Japanese American community, whereas originally before the redress [movement], I wouldn’t say that I was very much close to the community” (quoted in Takezawa, 1991, p. 48). at both direct and indirect experience with internment should yield lasting eects across generations is supported by other literature. General orientations toward politics such as partisan- ship and political cynicism are correlated across generations (Jennings and Niemi, 1968; Jennings, Stoker and Bowers, 2009). Moreover, as research on pre-adult political socialization has shown, intensive exposure to highly salient political events can crystallize aitudes toward policies and candidates, closing aitudinal gaps between adolescents and adults (Sears and Valentino, 1997). Furthermore, in the aermath of state violence and repression, group aachments, perceptions of victimhood, and feelings of threat are correlated across generations, despite a lack of direct experience with oppression among younger generations (Lupu and Peisakhin, 2017). For exam- ple, although conversations within families about internment were brief and infrequent, children still sensed the “racial implications of the internment” (Nagata and Cheng, 2003, 268). ere is also evidence in the U.S. that experience with the carceral state can have demobilizating eects on family members, albeit ones that are not long-lasting (White, 2019a). Taken together, the literature is in agreement that internment was, and is likely to be, an im- portant political event for those who experienced it as well as their family. However, in terms of impact, the literature suggests two divergent possibilities. First, in line with the comparative politics literature as well as the literature on Asian American public opinion, we expect that the internment experience had a galvanizing eect on Japanese-American political involvement. at is, internment was a seminal event that fomented a shared ethnic identity and raised Japanese- American political consciousness, leading to greater political engagement, activism, and political leadership. On the other hand, more in line with the literature on state-sponsored incarceration and its impact on American minority communities, we may expect that Japanese-American in-
8 ternment resonates more strongly with ndings on custodial citizenship, creating a depressive eect on subsequent political engagement among those aected and also their families. We may also expect that such a depressive eect could vary according to camp conditions, echoing theoriz- ing in the carceral eects literature. Ultimately, both of these are highly important: as the U.S. and other liberal democracies turn to internment in targeting largely minority migrant populations, scholars and policymakers must be aware of the possible downstream political consequences of these actions for these communities. We now turn to investigating these possible outcomes.
3 Japanese-American Internment Context and Data
Following the aack by Japanese air forces on Pearl Harbor, Hawaii, the United States entered World War II on December 8, 1941. On February 19, 1942, President Franklin Roosevelt, operat- ing with the encouragement of top military ocials, signed Executive Order 9066, authorizing the militarized internment of people of Japanese ancestry. us began the forcible relocation of more than 110,000 people of Japanese descent—the majority (62 percent) of whom were American citizens—into years-long internment into military camps. Public Proclamation 1, issued on March 2, 1942, by pro-internment U.S. Army General John L. Dewi, created two distinct military “exclusion zones.” Military Area 1 included swaths of the American west coast deemed sensitive for military purposes. is encompassed the coasts of California, Washington, and Oregon, as well as the southern sections of California and Arizona (Kashima et al., 2012). e remainder of these states constituted Military Area 2. Initially, all of those Japanese Americans living in Area 1 were under mandatory evacuation, but all California- based people of Japanese ancestry eventually fell under the military order. People of Japanese descent living further inland in Washington and Oregon, as well as those living away from the coast were not subject to evacuation or internment.6
6Hawaii, with a large Japanese-American population at the time, presents a very dierent case. While approxi- mately 2,000 Japanese and Japanese Americans living in Hawaii were interned in the continental U.S., the U.S. Army did not forcibly relocate the vast majority of Hawaii’s approximately 158,000 ethnically Japanese residents (Scheiber, Scheiber and Jones, 2009; Kashima, 2003). Instead, Hawaii’s governor declared martial law in 1941, placing the terri-
9 Figure 1: Historical map of the two military areas as well as the location of the Japanese-American internment camps. Source: Weglyn (1976).
us, as the historical map in Figure 1 shows, the eventual exclusion zone covered the entire state of California along with western sections of Washington, Oregon, and Nevada. Because people of Japanese ancestry disproportionately lived in California and other coastal areas, this had the eect of including the vast majority of people of Japanese ancestry in the exclusion zone. Specically, over 110,000—more than 86 percent—of Japanese Americans in the continental U.S. resided within the nal exclusion zone and were thus relocated from their homes; California alone contained around 73 percent of the total Japanese American population living in the continental U.S. (Ruggles et al., 2019).
tory under the command of the U.S. Army and subjecting the entire Hawaiian population to strict curfews and other restrictions. Because only a very small, targeted fraction—around 2 percent—of Hawaiians of Japanese descent were interned, we set Hawaii aside in our analyses. Additional qualitative context about Hawaiian internment is provided in Appendix Section 8.1. See also Kashima (2003, ch. 4).
10 Table 1: Internment Camp Characteristics
Camp State Peak Pop. Guard Towers Strikes Use of Force Violence Amache Colorado 7, 318 6 0 0 0 Crystal City Texas 4, 000 0 0 0 Jerome Arkansas 8, 497 7 0 0 1 Fort Sill Oklahoma 707 0 1 0 Heart Mountain Wyoming 10, 767 9 1 0 0 Minidoka Idaho 9, 397 8 0 0 0 Livingston Louisiana 1, 123 1 0 0 Lordsburg New Mexico 2, 500 1 1 0 Manzanar California 10, 046 8 0 1 1 Rohwer Arkansas 8, 475 8 0 0 0 Tule Lake California 18, 789 19 1 1 0 Poston Arizona 17, 814 0 1 0 1 Gila River Arizona 13, 348 1 0 0 0 Topaz Utah 8, 130 7 0 1 0 Sources: Ishizuka 2016, Burton, et. al. 2002, Densho Encyclopedia (see http://encyclopedia.densho.org/Fort Sill (detention facility)/), New Mexico Oce of the State Historian, (see http://newmexicohistory.org/places/lordsburg-internment-pow-camp) Note: Information on guard towers is not available for all facilities.
Internment Camp Site Selection. e U.S. Army began to scout locations that could col- lectively accommodate large concentrations of people in the spring of 1942. Sites suitable for internment had to be far from strategic military targets on the West Coast, large enough to house thousands of internees, and connected to transportation and public utilities. Because of these pressures, the U.S. Army focused on developing sites with these features already in place. Ac- cordingly, all but four assembly and relocation centers were located in pre-existing fairgrounds, stockyards, and exposition centers (Ng, 2002), all in remote locations. As we note above, a key motivation is to assess the extent to which camp-specic environ- ments impacted subsequent political engagement. Figure 1 shows the locations of the 10 major “War Relocation Centers,” (the largest internment camps) while Table 1 summarizes several camp characteristics. As the table shows, some camps had more military-style infrastructure, includ- ing watch towers and buildings dedicated to military use. We describe how these variables were coded in the data discussion, below.
11 Camp Assignment. e U.S. Army began evacuations on March 31, 1942 using a consistent procedure (Daniels, 1993). e larger military zones depicted in Figure 1 consisted of 108 Civil- ian Exclusion Zones encompassing approximately 1,000 people each. Army personnel posted signs displaying Civilian Exclusion Orders in public places (such as post oces, telephone poles, and store windows) throughout each of the 108 zones. Exclusion Orders informed people with Japanese ancestry (in practice, people who were at least 1/16 Japanese) that they were required to register themselves and their families and prepare for transfer. In the two days following, the head of each household would report to a control center near his or her home, register, and receive instructions for relocation. Evacuees were given six days to travel to one of multiple assembly centers located throughout California, Oregon, and Washington. Internees spent an average of 100 days in an assembly center before being transferred to a permanent internment camp (Kashima et al., 2012). Importantly for our analyses, with rare exceptions, evacuees were transferred from assembly centers to internment camps according to criteria unlikely to be correlated with their personal, cultural, or economic backgrounds. Evacuees in assembly centers with the most dangerous condi- tions (e.g., no indoor plumbing) were moved to internment camps rst (Burton, 2000). U.S. Army records suggest that it made eorts to move evacuees waiting at assembly centers to the nearest internment camps with climates closest to what they had known at home (Burton, 2000). Beyond these two concerns, families were assigned—together when possible—to internment camps that were (1) suciently complete in terms of construction to house evacuees and (2) had room for them. We see direct evidence of this in the War Relocation Authority’s records. Figure 2 shows that internees from Los Angeles County were primarily sent to the nearby assembly centers (le) in Manzanar, Pomona, and Santa Anita. e same internees were distributed across a variety of internment camps (right) in Arizona (Gila River, Poston), Colorado (Granada), Wyoming (Hearth Mountain), and Arkansas (Rowher, Jerome). us, conditional on initial pre-internment location, nal camp assignment was exogenous to family characteristics.
12 Figure 2: Distribution of Assembly Center and Internment Camp Locations among L.A. Internees
10000 10000
7500 7500
5000 5000 Internees Internees
2500 2500
0 0 None Tulare Mayer Topaz Fresno Turlock Poston Salinas Merced Jerome Rohwer Pomona Portland Puyallup Tanforan Pinedale Stockton Granada Minidoka Tule Lake Tule Manzanar Marysville Gila River Manzanar Santa Anita Sacramento Heart Mntn.
(a) Assembly Centers (b) Internment Camps
3.1 Data Sources and Key Variables
Our primary interest is in how the internment experience impacted the political behavior and aitudes of those interned and their direct descendants. For this, we draw on the Japanese Amer- ican Research Project (JARP), a nationally representative multi-wave survey of 4,153 mainland Japanese Americans that was conducted in 1967. (e survey excludes Hawaii-based people of Japanese descent. As we discuss above and in Appendix Section 8.1, this is not a problem for our substantive inferences as only a very small fraction of Hawaiians of Japanese ancestry were interned.) Although these data are older, we use JARP for several reasons. First, these data were collected at a time period when many of those who were interned were still alive and possibly politically active, thus allowing us to evaluate the political engagement of those who experi- enced internment rst-hand. (Making inferences about the direct internment experience using data gathered today would be impossible; as of our writing, those interned as very young infants
13 would be in their late 70s.) Second, other surveys of Asian Americans tend to be underpowered with regards to Japanese Americans. ird, JARP includes information about where respondents lived between 1932 and 1941, which allows us to leverage conditional exogenous variation in camp assignment (following Shoag and Carollo, 2016). Lastly, the study allows us to evaluate the eects of internment on political engagement across generations because it includes three immi- gration cohorts. ese are (1) Issei (Japan-born immigrants, or rst-generation immigrants from Japan), (2) Nisei (individuals born to Japan-born immigrants, or second generation), (3) Sansei (third generation). Given the cross-cohort nature of the survey, we selected outcomes that were present in at least two of the questionnaire forms (i.e., Issei, Nisei, or Sansei). Key features of the JARP data are summarized in Table 2. e JARP data are, however, constrained in that camp assignment is only available for Issei, despite the fact that many Nisei and even some Sansei were interned (Table 2). In the results that follow, we link respondents in JARP using the survey’s family identier eld and assign subse- quent generations the internment camp reported by Issei members. at is, we assume that all members of the same family were interned at the same place.7 In addition, in all of our main analyses, we include a xed eect for immigration generation (Issei, Nisei, or Sansei) since JARP used dierent questionnaires for each cohort. is also makes substantive sense: cultural and po- litical dierences across these immigration cohorts could have not only impacted the internment experience, but also inuenced political aitudes. In addition to JARP, we use the War Relocation Authority’s (“WRA”) records of Japanese internment in order to validate our results and provide richer descriptions of the internment ex- perience.8 e WRA and the Department of Justice recorded detailed personal information about internees, including name, age, gender, addresses prior to internment, family units, education,
7is assumption is plausible. First, the U.S. Army prioritized keeping families together. Second, families living close together were likely to be in the same Civilian Exclusion Zone and thus ordered to the same assembly centers. Assembly centers would oen send their whole populations to single internment camps when capacity was available. We recognize, however, that this assumption is potentially incorrect for some Nisei children living far enough from their parents to have been interned separately. 8See the Database of Japanese American Evacuees, Record Group 210, National Archives, https://www. archives.gov/research/JapaneseAmericans/wra.
14 Table 2: JARP Sample Demographic Information
Generation Observations Age Gender Married California Oregon Washington Other Interned Families Issei 1, 047 72 0.66 0.57 723 47 127 150 0.82 0.70 Nisei 2, 304 41 0.52 0.81 1, 470 101 266 467 0.76 0.86 Sansei 802 22 0.47 0.30 104 2 24 672 0.21 0.85 Total 4, 153 45 0.55 0.65 2, 297 150 417 1, 289 0.67 0.82 Notes: Age represents mean age; gender represents proportion male, and married represents proportion married. California, Oregon, Washington, and Other represent counts of Japanese Americans living in each location on the eve of the internment period. Interned represents the proportion of respondents in each generation who were interned Families represents the proportion respondents in a family where at least one member was interned and occupation. ese records contained assembly center and internment camp assignment in- formation for 109,384 Japanese Americans interned during WWII.
Internment Status. Given that the data span several immigration cohorts, we operationalized exposure to internment in three ways:
1. Direct Exposure: Respondents who were themselves interned, either solo or alongside mem- bers of their family
2. Family-Only Exposure: Respondents who were not themselves interned but had at least one family member in the sample who experienced internment9
3. Baseline: Respondents who were not interned and did not have an interned family member in the JARP.
In total, 67% of respondents in the JARP (across Issei, Nisei, and Sansei cohorts) fall into the rst category of direct exposure, 18% into the second category of family-only exposure, and 15% fall into the baseline or control group of no exposure whatsoever.
Individual covariates. We also include respondent gender and age, both recorded when re- spondents participated in the JARP survey waves. Age and gender are pre-treatment features that
9Each family in the JARP was assigned a family identier, which can be used to calculate how many respondents within a given family were interned.
15 may also aect our outcomes of interest. (Age, for instance, likely aects political engagement independently of the internment experience.) Finally, our analysis of camp-level eects relies on controls for respondents’ place of residence on the eve of internment. Since the internment camp to which people were ultimately conned was orthogonal to pre-treatment characteristics, it was largely a function of where they lived at the onset of WWII. All of our analyses of camp eects therefore control for pre-treatment area of residence.
Camp environment covariates. As we note above, part of our contribution lies in exploring whether the conditions of internment had potentially depressive eects, echoing what has been suggested by the carceral state literature (e.g., Weaver, Hacker and Wildeman, 2014). We do so by gathering new data on the conditions of the WRA camps themselves. First, we operationalize the level of militarism in camps by recording the numbers of watch towers per 1,000 people in each camp. Second, we also draw upon multiple historical resources to identify relocation cen- ters that experienced violence among internees, violence between internees and civilians, or the use of force by military personnel against internees. e strikes, violence, and force columns in Table 1 represent indicator variables for whether each event occurred in the camps. (“Vio- lence” represents violence between internees or violence between internees and the surrounding civilian population.) Lastly, we also incorporate other contextual features such as the percent of the surrounding county that was white (according to the 1940 U.S. Census) and the percent of the county-level vote-share won by Franklin Roosevelt in the 1940 presidential election. We use these camp characteristics in Section 6.1.
Outcome Variables. Our main outcomes of interest concern political engagement, a topic on which we have several JARP questions.10 First, “political interest” is measured using a four-item
10We do not present results on turnout and vote choice. Information on both items appears in the JARP, and our analysis suggests that internment experience has no detectable eect on either turnout or vote choice. However, both of these are expected outcomes given the previous literature describing internees. Regarding voter turnout, over 40% of Issei respondents in our data report not having U.S. citizenship. In addition, over 40% of Sansei in the JARP data were under 21 at the time the survey was administered, making them ineligible to vote. As a result, the JARP provides us with limited power to investigate this question. e same concerns apply to voter preference, but, in addition, the direction in which internment might aect vote choice between candidates from the two major parties is unclear.
16 ordinal scale ranging from “No interest at all” (0) to a “A great deal” (3) (x¯ = 1.29; s = .83).11 Second, “political engagement” is measured using a binary item asking respondents whether non-family members have asked them for advice regarding politics (pˆ = .17). ird, “faith in government” is measured using a binary item asking whether respondents disagreed that “most people in govern- ment are not really interested in problems of the average man.” (pˆ = .43), and nally, preferences for dissent are measured using a trichotomous variable that captures whether respondents would have preferred a leadership strategy that emphasized dissent (-1), accommodation (1), or neither (0) (x¯ = .61; s = .77). For this question, JARP asked whether Japanese-American leadership should have, in retro- spect, protested the circumstances of internment or cooperated with the U.S. government. Re- spondents could have answered that the best leadership strategy at the time would have been to protest internment (which we code using the value -1, for simplicity); they could have indicated that they were not sure what the best strategy was or that they couldn’t generalize on this point (0), or they could have claimed that the best leadership during the period would have sought a peaceful and orderly transition into internment facilities (1). Seventy-nine percent of respon- dents who were asked this question agreed that a peaceful and orderly leadership approach was preferable.
4 Eects of Internment Status
We rst investigate the possible role internment status—that is, the relationship between di- rect exposure to internment or family-only exposure to internment—played in shaping political engagement and political aitudes. For this, we leverage the fact that nearly 86 percent of peo- ple of Japanese descent living in the continental U.S. were residing in the exclusion zone, but 14 percent were not. Although not causal, the variation allows us to assess the relationship between
Internment was ordered by a Democratic president, but prominent Republicans had lobbied the federal government for restrictions on Japanese immigration and subsequent internment. 11e Issei version of this question had dierent response options, allowing for “No” or “Yes” responses. We code these responses as 0 and 1 on a 0-3 scale.
17 Internment Status and Political Engagement
Political Interest
Political Distrust
Group
Direct Exposure Family−Only Exposure Outcome
Political Advice
Leadership Approach
−0.3 −0.2 −0.1 0.0 0.1 0.2 0.3 Coefficient Estimate
Figure 3: e relationship between internment status and various indicators of political engage- ment, conditional on age, gender, generation, and pre-internment residential locations. A measure of political interest was present in all three generation forms, whereas the other three questions were only present in the Nisei and Sansei forms. 84% and 95% condence intervals are presented along with coecient estimates. 84% condence intervals (heavy bar) allow for visual tests of equality across coecients (Julious, 2004). Full model results are presented in the supplemental appendix. whether a person or one of their family members was sent to a camp and their subsequent political aitudes—a relationship that many scholars believe could have galvanized Japanese Americans’ political engagement (Wong, Ramakrishnan, Lee and Junn, 2011). Figure A.5 reports results from a linear model regressing indicators of political engagement on measures of direct exposure and family-only exposure to internment with pre-internment residential locations and generational identiers included as xed eects. (e full specication is provided in Appendix Table A.5.) e gure shows that those who were interned are about 13% of a scale point (± 9% of a scale point) less likely to report an interest in American politics than those
18 who were not, a statistically signicant dierence. ese paerns are similar among Japanese- Americans in our sample who themselves were not interned but who had family that were. ese individuals are about 18% of a scale point (± 9% of a scale point) less likely to express an interest in politics. ese estimates correspond to a movement of approximately 3% and 4% along a three- point scale, respectively. For both the distrust and political advice, estimates are in the expected direction, but there is considerable uncertainty. As for the leadership approach outcome, those who had direct exposure to internment are about 11% of a scale point (± 11% of a scale point) more likely to support a “peaceful and orderly” leadership approach during internment than one employing protest and dissent, relative to others. Among those who were not interned themselves but had family who were, this dierence is approximately 19% of a scale point (± 11% of a scale point). ese two estimates reect a 3% and 6% movement across a three-point scale, respectively. Consistent with an intergenerational transmission mechanism, coecient estimates for both measures of internment status are strikingly similar across outcomes. Formal tests of dierences between the two never reach conventional levels of statistical signicance (Appendix Table A.6).
Alternative Explanations in Military Service and Earning Potential. ese results pro- vide evidence that internment suppressed political interest. But could other explanations drive this nding? One possibility is that there is unexplained confounding between those of Japanese ancestry who were interned (or had family interned) and those that did not. However, using data from the 1940 Census, we fail to reject the null hypothesis of no mean dierence between Japanese Americans inside and outside of the exclusion zone with respect to gender, marital status, age, education, employment rates, and occupational class (see Appendix Section 8.7.2). What of other explanations stemming from the internment experience? We consider a likely possibility to be military service. During WWII, Japanese-American men of dra age were initially ineligible for military service. e U.S. government reversed this policy in 1943 and, ultimately, over 30,000 Japanese-Americans served in WWII. In this regard, studies have shown that those who served have higher levels of political participation (e.g., Meler, 2002; Teigen, 2006) and that there is an inter-generational component (Campante and Yanagizawa-Dro, 2015). us, an
19 alternative explanation is that those who were interned had lower rates of military service and this, not internment, drives our ndings. We investigate this by examining a JARP question that asked Nisei and Sansei respondents whether they (or their spouses) had ever served in the U.S. Armed Forces. While 55% of Nisei respondents and 34% of Sansei respondents reported serving in the military, we nd no signicant dierences between the baseline, direct exposure, and family-only exposure groups in terms of military service (see Appendix Table A.21). is suggests that Japanese-Americans who were interned are comparable to those who were not in terms of their willingness to serve in the U.S. military. Another possible explanation is that those interned suered nancially in ways that have persisted over time—perhaps as a result of lost professional opportunities or other lost income. Given that people who have fewer nancial resources tend to be less political engaged, this in- come dierential (and not the internment experience itself) could possibly explain our ndings. ough we nd dierences in total family income at the time of the JARP survey between the baseline group and those aected by internment (Appendix Table A.22), estimates for the direct exposure and family-only exposure group run in opposite directions, with higher incomes among those personally interned and lower incomes among those with family-only exposure. Given this inconsistency, and given the surprising positive nding among those directly interned, income dierentials are unlikely to explain the general paern of disengagement for both groups.
5 Eects of Internment Length
We next evaluate the association between internment length and political engagement. We would suspect, in accordance with the literature on ethnic targeting (Lupu and Peisakhin, 2017) and carceral contact in the United States (Weaver and Lerman, 2010), that longer interments would more strongly demobilize and depress civic engagement. Aer all, extremely brief detainments may have lile eect, but longer internments may expose internees to ongoing injustice, recal- ibrating their political views toward the U.S., and perhaps souring them on future engagement
20 Table 3: Internment Length - Summary Statistics
Internment Length N Percentage of Interned Sample Less than 1 Year 313 11.50 1 - 2 Years 498 18.30 2 - 3 Years 678 24.90 3 - 4 Years 1, 064 39.10 4 - 5 Years 157 5.80 5 + Years 8 0.30
(Weaver and Lerman, 2010). Table 3 provides summary statistics on internment length. Roughly 12 percent of internees in our sample were released within one year, while 45 percent were de- tained longer than three years. We operationalize this using simple ordinary least squares and subseing the data to only those with direct experience with internment—i.e., those who were themselves interned (either on their own or alongside family members). e data show that those interned for longer peri- ods had greater aenuation in political engagement, shown in Figure 4. (e full specication is provided in Appendix Table A.7). An additional year of being interned is associated with approx- imately 1.4% of a scale point decrease in political interest (± 2.6% of a scale point, so narrowly insignicant), a 4.2 pp increase in distrust (± 2.2 pp), a 3.4 pp decrease in the likelihood of being sought out for political advice (± 1.6 pp), and 4.3% of a scale point (± 3.4% of a scale point) increase in supporting a “peaceful and orderly” leadership approach during the internment process. To put this into context, those who were interned for four years or more (6% of the interned sample) are approximately 4% of a scale point less likely to report an interest in American politics than those who were interned for less than one year (12% of the interned subsample). Moreover, they are approximately 17 pp more likely to express distrust in government, 14 pp less likely to be sought out for political advice, and 17% of a scale point more likely to support a “peaceful and orderly” leadership strategy. is corresponds to a movement of about 6% across the the three-point scale.
21 Internment Length and Political Engagement
Political Interest
Political Distrust Outcome
Political Advice
Leadership Approach
−0.05 −0.02 0.00 0.03 0.05 0.08 Coefficient Estimate
Figure 4: e relationship between internment length and various indicators of political engage- ment, conditional on age, gender, and generation among those who were personally interned. A measure of political interest was present in all three generation forms, whereas the other three measures were only present in the Nisei and Sansei forms. 84% and 95% condence intervals are presented along with coecient estimates. Full model results are presented in the supplemental appendix.
6 Eects Associated With Internment Experience
ese ndings show broadly demobilizing eects associated with internment status and in- ternment length. In this section, we examine the internment experience more closely to try to understand the explanation behind these ndings. Not only does doing so shed light on how internment can lead to demobilization, but as as we explained in Section 3, conditional on ini- tial place of residence, assignment to one of the 10 major internment camps was unrelated to individual or family aributes (Shoag and Carollo, 2016). is allows us to estimate the causal ef- fects of exposure to specic camp conditions, conditional on internment, on downstream political
22 behavior in order to gauge possible mechanisms. We expect that these conditions are substantively important in shaping political aitudes. As suggested by the literature on custodial citizenship (Weaver, Hacker and Wildeman, 2014), in- ternees in camps that experienced more unrest, strikes, or backlash may have experienced greater demobilizing eects. First-person accounts of the internment experience frequently emphasize two socially pertinent features: (1) the struggle to access basic resources and (2) the reporting of (sometimes violent) widespread unrest among internees. On the rst point, shortages of ba- sic necessities were widespread. Flimsy living quarters meant that internees had lile protection from vermin and extreme weather (Pistol, 2017). Grievances over basic needs sparked strikes and demonstrations across several camps, inaming tensions among internee factions. At Tule Lake, agricultural workers protested authorities’ unwillingness to compensate the widow of a worker killed in a trucking accident; the camp’s project director responded by using Poston and Topaz internees as strike breakers (Burton, 2000). At Manzanar, union leader Harry Ueno led an investi- gation of supply shortages and founded the Mess Hall Workers Union in 1942. Tensions between Ueno’s union and the broadly pro-American Japanese-American Citizens League led to larger- scale violence among internees on multiple occasions (Burton, 2000). ese activities impacted large shares of aected camp populations. We expect that being imprisoned in a camp that witnessed strikes or violence among in- ternees might be demobilizing for several reasons. Internees who went on strike did so because they believed that camp authorities—the arm of the state with which they interacted most—were not commied to providing basic needs. At Manzanar, for example, Ueno’s commiee found ev- idence that shortages were caused in part by a camp ocial smuggling out supplies. Using the literature on the carceral state as a guide (Lerman and Weaver, 2014), experiences like this may have dampened faith in government and reduced the motivation to participate in politics. Vio- lent disagreements among internees would similarly have depressed political activity by straining communication and fostering resentments in the community. We explore these possible eects in Figures 5 (strikes) and 6 (violence). ese show the results
23 of OLS regressions of binary outcomes on whether the respondent was aliated with an intern- ment camp in which a strike or violent event took place. In both sets of analyses, we include xed eects for immigration cohort/survey wave (Issei, Nissei and Sansei), as well as controls for pre-internment location, age, and gender. (JARP includes case internment location only for Issei; we assume that interned Nisei and Sansei were sent to the same camps as their Issei relatives, an assumption we discuss above.) As before, we assess camp-treatment eects across two groups: (1) Direct Exposure (people who were themselves interned, either solo or alongside family) or (2) Family-Only Exposure (people who were not interned, but had Issei family members who were). We do not estimate eects on those who were not interned and had no family interned.
Strikes. We rst examine the causal eect of being assigned to a camp that experienced strikes. Figure 5 shows that respondents who were themselves interned in camps where strikes took place were 5% of a scale point (± 8% of a scale point) less likely to report being interested in American politics, although this is narrowly insignicant. Among those who were not themselves interned, but had family members who were, this treatment eect is of a larger magnitude (nearly 9% of a scale point), but insignicant (± 11% of a scale point). In terms of trust in government, respondents who were interned themselves were approximately 8 pp (± 5 pp) more likely to say that the government is not concerned with the issues facing everyday people; results are similar for respondents who only experience internment via family. Both are signicant. As for political advice, respondents interned at camps that witnessed strikes were 2.4 pp (± 2 pp) less likely to report having been approached for political advice, though this nding is signicant only for those who themselves were interned (and not for those who only had family interned). Lastly, the results suggest that internees sent to camps that experienced strikes were signicantly more likely to favor leaders who espoused peaceful transitions, perhaps because such respondents did not believe protesting would yield concessions or because they feared retribution. On this point, those directly interned are 0.08 (± 0.05) scale points (relative to the -1, 0, 1 scale) more likely to prefer leaders who fostered orderly transitions. e ndings are similar are roughly similar in magnitude but insignicant for those who were not themselves interned but who had
24 family who were. Overall, the results lend support to the hypothesis that labor unrest and relative deprivation at camps has a demobilizing eect on internees. We mostly detect slightly stronger ndings for those individuals themselves interned, although formal tests looking at the dierence in treatment eect between direct versus family-only exposure are insignicant. (Appendix 8.6). is provides suggestive evidence in favor of an inter-generational transmission of aitudes, consistent with our ndings above.
Violence. We next examine the dierences in outcomes in terms of whether internees were assigned to camps that witnessed violence. ese results are presented in Figure 6. In terms of political interest, respondents who lived through violent episodes while they were interned express signicantly lower levels of interest in American politics than counterparts who did not experience violence. ese respondents were 7.5% of a scale point (± 6% of a scale point) less likely to report an interest in American politics. However, the eect is only signicant for those who themselves were interned, either on their own or alongside family. For individuals who only experienced the eects of internment via interned family, the eect is in a similar direction but not signicant. For trust in government, internees imprisoned in camps that witnessed violent episodes amongst internees themselves were 2 pp (± 7 pp) more likely to report believing that the government had lile concern for the problems faced by average people. Respondents who only had fam- ily interned were signicantly more likely to express skepticism of the government than other internees. is group was 10pp (± 4pp) more likely to report high levels of distrust. Respondents interned in camps that experienced violence among internees also reported lower rates (2 pp) of being sought out for political advice, but the nding is not signicant ei- ther for those directly interned or those who only had family interned. Lastly, the respondents directly interned in camps that experienced violence were 0.08 scale points (± 0.07) more likely to support leaders who did not favor protest. e eect for those were not themselves interned but who had family members who were are similar in magnitude, but narrowly insignicant.
25 Effect of Witnessing Strikes While Interned
● Political Interest ●
● Political Distrust ●
Group
● Direct Exposure ● Family−Only Exposure Outcome
● Political Advice ●
● Leadership Approach ●
−0.2 −0.1 0.0 0.1 Coefficient Estimate
Figure 5: Eects associated with strike conditions in a camp. 95% (narrow bar) and 84% (heavy bar) condence intervals are shown. Controls include age, gender, pre-internment residential location, and survey wave (Issei, Nisei, Sansei). Political interest models rely on data from all three generations; political advice, leadership approach, and political distrust models are based on the Nisei and Sansei waves. Sample sizes are reported in Table A.9 of the supplemental appendix.
Overall, these eects are slightly more modest than those we found regarding strikes. How- ever, they provide suggestive evidence that violence had a demobilizing eect. In addition, we cannot rule out dierences in treatment eects transmied across generations through family experiences (shown in Appendix 8.6), although our ndings are mostly signicant for those who experienced internment themselves.
26 Effect of Witnessing Violence While Interned
● Political Interest ●
● Political Distrust ●
Group
● Direct Exposure ● Family−Only Exposure Outcome
● Political Advice ●
● Leadership Approach ●
−0.1 0.0 0.1 0.2 Coefficient Estimate
Figure 6: Eects associated with violent conditions in a camp. 95% (narrow bar) and 84% (heavy bar) condence intervals are shown. Controls include age, gender, pre-internment residential location, and survey wave (Issei, Nisei, Sansei). Political interest models rely on data from all three generations; political advice, leadership approach, and political distrust models are based on the Nisei and Sansei waves. Sample sizes are reported in Table A.8 of the supplemental appendix.
6.1 Robustness of the Results
In sum, our results lend support to a broad demobilizing eect associated with internment, with particularly strong eects among those who were interned longer and those with direct connections to camps exhibiting labor conict or violence. We also nd a corresponding, albeit slightly weaker eect among those with no direct internment experience but who had family
27 members interned, lending evidence to the possibility of inter-generational transmission of ai- tudes. To this extent, our ndings not only support the broader literature on custodial citizenship in the U.S. but expand it by conrming that internment length and conditions are an important pathway in the demobilizing eect. We note, however, some threats to this analysis. ese are (1) confounding in camp assignment and (2) dierences in pre-internment locations.
Robustness to Unobserved Confounders in Camp Assignment. Relying on the historical record and following other work on internment (e.g., Shoag and Carollo, 2016), we assume that, for those individuals who were themselves interned (either solo or alongside family members), camp assignment is orthogonal to pre-internment characteristics at the individual level. However, in- ternees may dier on important characteristics impacting political engagement, even conditional on location. One possible threat to our strategy is the possibility that the government reassigned or se- questered internees based on their level of resistance aer their initial camp assignment. An example of this is the WRA’s loyalty questionnaire, which was administered to men aged 17 and over beginning in 1943. Respondents who answered that they would not register for selective service and could not pledge unconditional loyalty to the U.S. were labeled “disloyal” and se- questered at the Tule Lake camp. We address possible confounding by replicating the analyses subseing to JARP respondents who were younger than 17 in 1945. is group would not have been subjected to the questionnaire and is less likely to have expressed other resistance owing to their age. e results from that analysis are consistent with the eects of internment that we report in this section. (See Appendix Figures A.15 through A.18 for these results.) We also assess whether camp assignment is related to any pre-treatment controls (such as age, gender, or occupation) that could also inuence political engagement. We do so only for those who themselves were interned. Since internment camps are discrete units, we model camp as- signment as a multinomial logistic function of age, gender, and pre-internment location—the most relevant pre-treatment control variables we have access to in JARP. Values in Table 4 correspond
28 to z-scores, calculated by dividing coecients from the multinomial logit by their corresponding standard errors, for each covariate. In the age and gender columns, no camp is associated with a test statistic greater than 1.96 or less than -1.96. (Pre-internment location, on the other hand, is signicantly associated with camp assignment in our model, but this is what we expect given the way camp assignment was carried out.12)
Table 4: Covariate Balance Across Internment Camps
Age Gender (Male) Pre-Camp Location
Jerome -.88 1.27 -1.43 Heart Mountain .38 -.54 -2.74 Minidoka .14 .31 -1.97 Manzanar -.18 1.58 -2.73 Rohwer 1.37 .15 -4.17 Tule Lake -.49 .68 -1.07 Poston -.09 .12 -2.06 Gila River .04 -.64 -1.41 Topaz .67 -.33 2.34
Note: Reference category is Granada (Amache)
Finally, plots of age, gender, and pre-internment occupations for internees held at dierent camps (displayed in Appendix Figures A.9 and A.10 and Appendix Table A.14) reveal no signicant dierences in these distributions across dierent internment locations. is is conrmed in a series of Kolmogorov-Smirnov (KS) tests comparing the distributions, presented in the Appendix. ese provide evidence that internee camp assignment was a function of location (per Table 4) but not pre-treatment characteristics like age and gender. is gives us good reason to think that
12ese results only include internment camps to which at least 30 respondents in the JARP reported being assigned; estimates for camps with fewer assigned respondents are unlikely to be reliable. In the JARP data, pre-camp location is a three-digit numeric code, which we leave as a numeric variable in this model to preserve power; locations in the same region and state will be close in value and locations across states will dier considerably in value, which provides reasonable distinctions between dierent pre-camp locations.
29 Figure 7: Assignment to Internment Camp by State of Residence
California Oregon Washington California Oregon Washington 400
15000 300
10000 200 Internees Internees
5000 100
0 0 Topaz Topaz Topaz Topaz Topaz Topaz Poston Poston Poston Poston Poston Poston Jerome Jerome Jerome Jerome Jerome Jerome Rohwer Rohwer Rohwer Rohwer Rohwer Rohwer Granada Granada Granada Granada Granada Granada Minidoka Minidoka Minidoka Minidoka Minidoka Minidoka Tule Lake Tule Lake Tule Lake Tule Lake Tule Lake Tule Lake Tule Gila River Gila River Gila River Gila River Gila River Gila River Manzanar Manzanar Manzanar Manzanar Manzanar Manzanar Heart Mountain Heart Mountain Heart Mountain Heart Mountain Heart Mountain Heart Mountain
(a) WRA Data (b) JARP Data assignment was also orthogonal to other pre-treatment traits.
Location Prior to Internment. Identifying the eect of camp environment relies on the as- sumption that every person of Japanese ancestry who lived in roughly the same area on the eve of internment was treated similarly by the WRA, which prioritized rapid evacuation and proxim- ity. Figure 7 provides evidence that supports this assignment mechanism. Each panel in Figure 7 displays the distribution of internees living in California, Oregon, and Washington on the eve of WWII. Panel (a) describes this information using the War Relocation Authority’s records, while Panel (b) shows the same distribution using JARP data. is is again, only for those individuals who themselves were interned. If the assumption that the U.S. Army prioritized proximity and rapid transportation from assembly centers to camps holds, then most internees from a given state should have been held in camps closest to their state of residence. is is borne out by the two gures. For example, Japanese Americans living in Oregon and
30 Washington were sent to the northern-most camps in California (Tule Lake), Wyoming (Heart Mountain), and Idaho (Minidoka). ose living in California were sent to camps located near them—these included Manzanar, Poston, and Gila River for people living in southern California and Tule Lake for people living in northern California. Lastly, we note strong correspondence between the two gures, giving us assurance that JARP self-reported data paints an accurate portrayal of internment assignment.
6.2 Alternative Mechanisms of Demobilization
Our analyses suggest that dierences in political engagement among Japanese Americans var- ied not just by internment status, but also by features of the camps themselves, with exposure to unrest being an important mechanism behind demobilization. However, one potential challenge to this substantive interpretation is the possibility that any camp eects are actually the result of another factor correlated with internment location. Here, we examine the three likeliest al- ternative factors: (1) use of force by the state, (2) a militarized environment, and (3) the camps’ surrounding racial and political environments.
Use of Force By the State. Previous work on the carceral state implies that the use of force by the state would have politically demobilizing eects via instilling fear of the state, which could possibly explain our earlier ndings. However, this prediction is not borne out, as shown in Figure 8b, which analyzes camp eects according to whether the camp experienced at least one instance use of force by military personnel against internees. e point estimates of this eect on political distrust, advice, interest and preferences over leadership are close to zero. Only among those with family exposure only is a single signicant nding (on political interest); all other ndings are insignicant. ese non-ndings are at rst glance puzzling. One reason we might observe demobilizing eects for violence among internees (discussed in our previous section), but not for cases in which military personnel used force against internees, is the dierence in scale. Episodes of violence
31 Figure 8: Camp Eects: Militarized Conditions
Effect of Being Interned in Highly Militarized Camp Effect of Witnessing Use of Military Force While Interned
● ● Political Political Interest Interest ● ●
● ● Political Political Distrust Distrust ● ●
Group Group
● Direct Exposure ● Direct Exposure ● Family−Only Exposure ● Family−Only Exposure Outcome Outcome
● ● Political Political Advice Advice ● ●
● ● Leadership Leadership Approach Approach ● ●
−0.1 0.0 0.1 0.2 −0.1 0.0 0.1 Coefficient Estimate Coefficient Estimate (a) Militarism (b) Use of Force Note: 84% and 95% condence intervals are shown. Controls include age, gender, and pre- internment residential locations. Political interest models rely on data from all three generations; political advice, leadership approach, and political distrust models are based on the Nisei and Sansei sample. among internees generally tended to precede or follow large demonstrations or involve large groups. Incidents in which guards used force against internees, in contrast, tended to be more isolated and rare. To give one example, one such incident coded in our data involves a guard at Topaz camp fatally shooting an elderly internee for standing too close to a perimeter fence (Burton, 2000). Violent confrontations between large groups of internees would have exposed more internees to unrest than isolated shootings such as this, which involved few individuals.
Militarized Space. Another jarring feature of the internment experience was living under mili- tarized conditions. ese conditions manifested in physical space through the use of guard towers, barbed wire fencing, and barracks-style housing for internees. ese physical elements served as “reminders of [internees’] lack of freedom” (Burton, 2000). We operationalize the militarization of space as the number of guard towers per one thousand internees at peak camp population. e results of an analysis looking at the eect of these on
32 political engagement appear in Figure 8a. Respondents interned in more militarized camps (or with relatives who were) were 8% of a scale point (± 8% of a scale point) more likely to express interest in American politics, a nding in the opposite direction from what we would expect. is eect is, however, only signicant respondents who themselves were not interned but who had family members who were. Other ndings are, across the board, insignicant and with point estimates close to zero. is suggest that the demobilization paerns observed in the JARP sample is due to the internees’ collective exposure to unrest, rather than dierences in the camps’ physical features.
Surrounding Political and Racial Environment. Although camp locations were remote and internees had limited contact with civilians living nearby, a possibility is that exposure to local culture or local political aitudes provide a possible alternative conduit for the long term eects we report in Section 6, particularly for those who themselves were interned. For example, if internees were sent to camps located in nearly exclusively white areas, this might have led them to have lowered feelings of belonging, thereby suppressing overall political engagement. Similarly, internment in camps located in extremely conservative areas might have had a depressing impact on political engagement in a what is a more liberal-leaning group (Wong, Ramakrishnan, Lee and Junn, 2011). We test for this possible alternative explanation by analyzing the eects of (1) the proportion of the internment camp’s county population that was white in 1940 and (2) the political climate, as measured by Franklin Roosevelt’s 1940 share of the two-party vote. To the rst point, Appendix Tables A.17 through A.20 present the results of models that include the percent of the camp’s county that was white and the percentage of the two-party vote received by Franklin Roosevelt in 1940 as the primary explanatory factors. As the analyses show, county demographics do not consistently have a signicant eect on the various political engagement outcomes (showing a signicant relationship in only two out of 16 specications), meaning that this is unlikely driver behind our results. (Including this factor as a control in the same model as camp strikes and violent episodes does not change either the magnitude or signicance of those ndings, as shown
33 in Tables A.17 and A.18.) For local Democratic-party vote share, we see mixed ndings, with some evidence that area conservativeness is negatively liked with some political engagement (for example on political advice); however, the results are not consistent across groups or outcomes. In addition, including Democratic-party vote share as a control in the same model as strikes or violence does not change the fact that those are statistically and substantively signicant. e overall ndings are consistent with the interpretation we presented in Section 6, which is that the interpersonal, on-the-ground environment was important in terms of the overall depressive eect of internment.
7 Conclusion
Aer Pearl Harbor, raw racism and xenophobia led over a hundred thousand people of Japanese ancestry to be transported to militarized internment camps, where they were held—against their will and unconstitutionally—for years over the course of WWII. In this study, we explored the downstream consequences of this seminal, racist event. Did the targeting by the state of one minority group have downstream consequences on that group’s political aitudes and political engagement? Our ndings suggest that the internment of people of Japanese ancestry did indeed impact their subsequent political behavior, in ways that speak not just to scholarly discussions but also to ongoing current events. First, we examine the eects of such internment by comparing po- litical engagement among those who were themselves interned (or had family interned) to those who were not. Conditional on pre-internment residential location, we nd that those with di- rect and family-only experience with internment are less likely to express an interest in politics and prefer a confrontational strategy toward the government relative to those who were not in- terned. Second, we explored the impact of internment itself, conditional on people being interned. Leveraging the fact that camp assignment was exogenous to individual or family characteristics conditional on pre-internment location, we nd that internment length and camp conditions are associated with decreases in political engagement. is suggests that longer and more hostile
34 contacts with the state contributed to larger demobilization eects among those connected to the experience. Lastly, across every measure of political engagement, we cannot nd meaningful dif- ferences in the eect of internment among those directly interned versus those who experience internment through a family member. is suggests a strong depressing eect of internment on Japanese Americans’ political engagement that extend to even those with more indirect exposure. Our study makes several contributions to ongoing scholarly discussions. From the perspec- tive of the comparative politics literature, our paper highlights the potential demobilizing eects of punitive interactions with the state in the unusual context of a liberal democracy. is is an important inquiry: to date, existing work on the topic of repression, violence, and ethnic targeting has mostly focused on weak states, with more recent studies considering the eects of author- itarian strong states such as the Soviet Union during Stalin’s reign. Our ndings dovetail with those of Lupu and Peisakhin (2017) and Rozenas, Schue and Zhukov (2017) who nd that repres- sion and violence increases distrust toward the state. (However, in contrast to Lupu and Peisakhin (2017), we do not observe increased mobilization and political engagement among members of the targeted group.) Future research could explore whether these dierences between democracies and autocratic regimes are shaped by institutional variation or the nature of repression eorts. We also contribute to research in American politics examining the political consequences of growth in the carceral state. While much of this growing literature has viewed exposure to pe- nal institutions in a binary fashion (contact versus no contact), we nd evidence that conditions within these institutions maer, such that harsher conditions increase the likelihood of demo- bilization, distrust, and disengagement. Potential avenues for future research research on the “carceral state” could assess whether variation in the severity and duration of punitive encoun- ters with the state yield more dire consequences for aected groups. Lastly, our paper contributes to an important and growing literature on Asian-American pub- lic opinion and political behavior. Previous work has demonstrated that Japanese Americans have high rates of political participation and leadership, with scholars positing that the group’s pre- vious history of internment may have had a galvanizing eect (Wong, Ramakrishnan, Lee and
35 Junn, 2011). Our ndings call this laer explanation into question. Indeed, our ndings sug- gest that those with intimate or secondary contact with internment are, if anything, less likely to be engaged, trustful of the government, and politically participatory. is leaves open sev- eral questions that are worthy of further research, including whether internment indirectly had galvanizing eects via the movement for redress in the 1970s and 80s (Takezawa, 1991). We conclude by noting that our ndings have high relevance beyond scholarly discourse. roughout liberal democracies, governments are increasingly turning to the detention of mi- nority groups as a public policy (Sampson and Mitchell, 2013). is includes state-sponsored detentions of migrants, group-based repatriation, refugee camps, and carceral policies that dis- proportionately impact members of minority groups. Our study—which suggests that detentions that happen in the modern-day could suppress political engagement among these groups for gen- erations to come—urges caution in this approach.
36 References
Arendt, Hannah. 1973. e Origins of Totalitarianism. Houghton Miin Harcourt.
Bellows, John and Edward Miguel. 2009. “War and Local Collective Action in Sierra Leone.” Jour- nal of Public Economics 93(11-12):1144–1157.
Blaman, Christopher. 2009. “From Violence to Voting: War and Political Participation in Uganda.” American Political Science Review 103(2):231–247.
Bosworth, Mary and Sarah Turnbull. 2014. “Immigration Detention, Punishment, and the Crimi- nalization of Migration.” e Routledge Handbook on Crime and International Migration pp. 91– 106.
Bowler, Shaun, Stephen P. Nicholson and Gary M. Segura. 2006. “Earthquakes and Aershocks: Race, Direct Democracy, and Partisan Change.” American Journal of Political Science 50(1):146– 159.
Burch, Traci. 2013. Trading Democracy for Justice: Criminal Convictions and the Decline of Neigh- borhood Political Participation. University of Chicago Press.
Burton, Jeery F. 2000. Connement and Ethnicity: An Overview of World War II Japanese Amer- ican Relocation Sites. Publications in Anthropology (Western Archeological and Conservation Center (U.S.)) ; no. 74 Tucson, Ariz.]: Western Archeological and Conservation Center, National Park Service, U.S. Dept. of the Interior.
Campante, Filipe and David Yanagizawa-Dro. 2015. “e Intergenerational Transmission of War.” National Bureau of Economic Research Working Paper.
Daniels, Roger. 1993. Prisoners without Trial: Japanese Americans in World War II. New York: Hill and Wang.
Davenport, Christian. 2005. “Repression and Mobilization: Insights from Political Science and Sociology.” Repression and Mobilization pp. vii–xli.
Dehdari, Sirus H and Kai Gehring. 2018. “e Origins of Common Identity: Evidence from Alsace- Lorraine.”.
Epp, Charles R., Steven Maynard-Moody and Donald P. Haider-Markel. 2014. Pulled Over: How Police Stops Dene Race and Citizenship. University of Chicago Press.
Epstein, Jerey H. 2016. Democracy and Its Others. Bloomsbury Publishing.
Fares, Melissa. 2017. “75 Years Later, Japanese Americans Recall Pain of Internment Camps.” Reuters.
37 Fouka, Vasiliki. 2019. “How do Immigrants Respond to Discrimination? e Case of Germans in the U.S. during World War I.” American Political Science Review 113(2):405–422.
Hjalmarsson, Randi and Mark Lopez. 2010. “e Voting Behavior of Young Disenfranchised Felons: Would ey Vote if ey Could?” American Law and Economics Review 12(2):356–393.
Homola, Jonathan, Migue Pereira and Margit Tavits. 2020. “Legacies of the ird Reich: Concen- tration Camps and Outgroup Intolerance.” American Political Science Review .
Jennings, M Kent, Laura Stoker and Jake Bowers. 2009. “Politics Across Generations: Family Transmission Reexamined.” e Journal of Politics 71(3):782–799.
Jennings, M Kent and Richard G. Niemi. 1968. “e Transmission of Political Values from Parent to Child.” American Political Science Review 62(1):169–184.
Julious, Steven A. 2004. “Using Condence Intervals Around Individual Means to Assess Statistical Signicance Between Two Means.” Pharmaceutical Statistics: e Journal of Applied Statistics in the Pharmaceutical Industry 3(3):217–222.
Kalhan, Anil. 2013. “Immigration Policing and Federalism rough the Lens of Technology, Surveillance, and Privacy.” Ohio State Law Journal 74:1105.
Kashima, Tetsuden. 2003. Judgment Without Trial: Japanese American Imprisonment During World War II. University of Washington Press.
Kashima, Tetsuden et al. 2012. Personal Justice Denied: Report of the Commission on Wartime Relocation and Internment of Civilians. University of Washington Press.
Kupchik, Aaron and omas J. Catlaw. 2015. “Discipline and Participation: e Long-term Eects of Suspension and School Security on the Political and Civic Engagement of Youth.” Youth & Society 47(1):95–124.
LeBas, Adrienne. 2006. “Polarization as cra: Party formation and state violence in Zimbabwe.” Comparative Politics 38(4):419–438.
Lee, Hedwig, Lauren C. Porter and Megan Comfort. 2014. “Consequences of Family Member Incarceration: Impacts on Civic Participation and Perceptions of the Legitimacy and Fairness of Government.” e ANNALS of the American Academy of Political and Social Science 651(1):44– 73.
Lerman, Amy E. and Vesla M. Weaver. 2014. Arresting Citizenship: e Democratic Consequences of American Crime Control. University of Chicago Press.
Levy, Jack S. 1988. “Domestic Politics and War.” e Journal of Interdisciplinary History 18(4):653– 673.
38 Lupu, Noam and Leonid Peisakhin. 2017. “e Legacy of Political Violence Across Generations.” American Journal of Political Science 61(4):836–851.
Lyall, Jason. 2009. “Does Indiscriminate Violence Incite Insurgent Aacks? Evidence From Chech- nya.” Journal of Conict Resolution 53(3):331–362.
Maltby, Elizabeth. 2017. “e Political Origins of Racial Inequality.” Political Research arterly 70(3):535–548.
Mann, Michael. 2005. e Dark Side of Democracy: Explaining Ethnic Cleansing. Cambridge Uni- versity Press.
McGarry, John. 1998. “’Demographic Engineering’: e State-directed Movement of Ethnic Groups as a Technique of Conict Regulation.” Ethnic and Racial Studies 21(4):613–638.
Meler, Suzanne. 2002. “Bringing the state back in to civic engagement: Policy feedback eects of the GI Bill for World War II veterans.” American Political Science Review 96(2):351–365.
Mickey, Robert. 2015. Paths Out of Dixie: e Democratization of Authoritarian Enclaves in Amer- ica’s Deep South, 1944-1972. Princeton University Press.
Nagata, Donna K. and Wendy J.Y. Cheng. 2003. “Intergenerational Communication of Race- Related Trauma by Japanese American Former Internees.” American Journal of Orthopsychiatry 73(3):266–278.
Nanwani, Shaira. 2011. “e Burqa Ban: An Unreasonable Limitation on Religious Freedom or a Justiable Restriction.” Emory International Law Review 25(3):1431–1475.
Ng, Wendy. 2002. Japanese American Internment During World War II: A History and Reference Guide. Greenwood Press.
Pantoja, Adrian D., Ricardo Ramirez and Gary M. Segura. 2001. “Citizens by Choice, Voters by Ne- cessity: Paerns in Political Mobilization by Naturalized Latinos.” Political Research arterly 54(4):729–750.
Pistol, Rachel. 2017. Internment During the Second World War : A Comparative Study of Great Britain and the USA. Bloomsbury Publishing.
Radack, Jesselyn A. 2004. “You Say Defendant, I Say Combatant: Opportunistic Treatment of Terrorism Suspects Held in the United States and the Need for Due Process.” N.Y.U. Review of Law & Social Change 29(3):525–553.
Reeves, Richard. 2015. Infamy: e Shocking Story of the Japanese American Internment in World War II. Henry Holt and Company.
39 Roche, Jean-Jacques. 2014. “Walls and Borders in a Globalized World: e Paradoxical Revenge of Territorialization.” Borders, Fences and Walls: State of Insecurity pp. 105–115.
Rozenas, Arturas, Sebastian Schue and Yuri Zhukov. 2017. “e Political Legacy of Violence: e Long-Term Impact of Stalin’s Repression in Ukraine.” e Journal of Politics 79(4):1147–1161.
Ruggles, Steven, Sarah Flood, Ronald Goeken, Josiah Grover, Erin Meyer, Jose Pacas and Mahew Sobek. 2019. “Integrated public use microdata series: Version 9.0 [Machine-readable database].” Minneapolis: University of Minnesota .
Sampson, Robyn and Grant Mitchell. 2013. “Global Trends in Immigration Detention and Alter- natives to Detention: Practical, Political and Symbolic Rationales.” Journal on Migration and Human Security 1(3):97–121.
Scheiber, Harry N., Jane L. Scheiber and Benjamin Jones. 2009. “Hawaii’s Kibei Under Martial Law: A Hidden Chapter in the History of World War II Internments.” Western Legal History 22(1 & 2):1–102.
Schnapper, Dominique, Chantal Bordes-Benayoun and Freddy Raphael. 2011. Jewish Citizenship in France: e Temptation of Being Among One’s Own. Transaction Publishers.
Sears, David O. and Nicholas A. Valentino. 1997. “Politics Maers: Political Events as Catalysts for Preadult Socialization.” American Political Science Review 91(1):45–65.
Shoag, Daniel and Nicholas Carollo. 2016. “e Causal Eect of Place: Evidence from Japanese- American Internment.” Working Paper. URL: https://scholar.harvard.edu/files/shoag/files/ internment working paper.pdf
Soss, Joe and Sanford F. Schram. 2007. “A Public Transformed? Welfare Reform as Policy Feed- back.” American Political Science Review 101(1):111–127.
Soss, Joe and Vesla Weaver. 2017. “Police Are Our Government: Politics, Political Science, and the Policing of Race–Class Subjugated Communities.” Annual Review of Political Science 20:565–591.
Takezawa, Yasuko I. 1991. “Children of Inmates: e Eects of the Redress Movement Among ird Generation Japanese Americans.” alitative Sociology 14(1):39–56.
Teigen, Jeremy M. 2006. “Enduring Eects of the Uniform: Previous Military Experience and Voting Turnout.” Political Research arterly 59(4):601–607.
Uggen, Christopher and Je Manza. 2002. “Democratic Contraction? Political Consequences of Felon Disenfranchisement in the United States.” American Sociological Review 67(6):777–803.
Verba, Sidney, Kay Lehman Schlozman and Henry E. Brady. 1995. Voice and Equality: Civic Vol- untarism in American Politics. Cambridge, Mass: Harvard University Press.
40 Wahab Twibell, Ty S. 2004. “e Road to Internment: Special Registration and Other Human Rights Violations of Arabs and Muslims in the United States.” Vermont Law Review 29(2):407– 553.
Walker, Hannah L. 2014. “Extending the eects of the carceral state: Proximal contact, political participation, and race.” Political Research arterly 67(4):809–822.
Weaver, Vesla M. and Amy E. Lerman. 2010. “Political Consequences of the Carceral State.” Amer- ican Political Science Review 104(4):817–833.
Weaver, Vesla M., Jacob S. Hacker and Christopher Wildeman. 2014. “Detaining Democracy? Criminal Justice and American Civic Life.” 651(1):6–21.
Weglyn, Michi. 1976. Years of Infamy: e Untold Story of America’s Concentration Camps. William Morrow & Co.
Welch, Michael. 2002. Detained: Immigration Laws and the Expanding INS Jail Complex. Temple University Press.
White, Ariel. 2016. “When reat Mobilizes: Immigration Enforcement and Latino Voter Turnout.” Political Behavior 38(2):355–382.
White, Ariel. 2019a. “Family Maers? Voting Behavior in Households with Criminal Justice Contact.” American Political Science Review 113(2):607–613.
White, Ariel. 2019b. “Misdemeanor Disenfranchisement? e Demobilizing Eects of Brief Jail Spells on Potential Voters.” American Political Science Review 113(2):311–324.
Wong, Janelle, S Karthick Ramakrishnan, Taeku Lee and Jane Junn. 2011. Asian American Political Participation: Emerging Constituents and eir Political Identities. Russell Sage Foundation.
Wong, Janelle S. et al. 2011. Asian American Political Participation: Emerging Constituents and eir Political Identities. Russell Sage Foundation.
WRA, War Relocation Authority. 1942. Second arterly Report. Harry S. Truman Library Col- lections.
WRA, War Relocation Authority. 1943. News Release: Work of the War Relocation Authority, An Anniversary Statement by Dillon S. Myer. Harry S. Truman Library Collections.
Zakaria, Fareed. 1997. “e Rise of Illiberal Democracy.” Foreign Aairs pp. 22–43.
Zhukov, Yuri M. and Roya Talibova. 2018. “Stalin’s Terror and the Long-term Political Eects of Mass Repression.” Journal of Peace Research 55(2):267–283.
41 8 Appendix
8.1 People of Japanese Descent Living in Hawaii
Our analysis excludes the large number of people of Japanese descent living in Hawaii at the time. We exclude these individuals on substantive grounds, which we further explain in this section.13 During the mid-twentieth century, substantial numbers of people of Japanese descent lived in Hawaii, reecting the close historical contact between Hawaii and Japan. In the late 1930s and early 1940s, this population included approximately 158,000 people of Japanese descent (citizens and non-citizens), accounting for around 37 percent of the territory’s population. By contrast, there were around 130,000 people of Japanese descent living in the entire mainland U.S., much fewer in terms of the absolute number and also as a share of the population. Given the large shares of Japanese and Japanese-Americans in Hawaii, large-scale evacua- tion and internment would be dicult. Some individuals in the military and in the government nonetheless recommended removing all Issei and their citizen children to the mainland or taking them hostage (Kashima, 2003, p. 68); however, others with deeper knowledge of the situation in Hawaii—such as Delos Emmons, head of Western Defense Command during the war—warned that removing such a large population would prove logistically intractable and leave gaping needs in terms of much needed war-time labor (Kashima, 2003, p. 75). e laer view eventually won out. Following the aack by the Japanese on Pearl Harbor on December 7, 1941, the entirety of Hawaii was placed under martial law, with control of power transferred from the civilian territorial governor the military head of the Hawaii command. (is lasted some three years, until October of 1944.) e introduction of martial law aected all resi- dents of Hawaii, including citizens and non-citizens, Japanese-Americans and others. However, there were some detentions and evacuations. For example, several hundred Japanese or Japanese Americans on a special FBI “custodial detention list” were quickly taken into custody.
13For more on the dierences in the internment of Hawaii-based people of Japanese descent and those living on the mainland, see Kashima (2003, ch. 4).
1 is included “businesspeople, Japanese consular ocials, Japanese language school teachers and principals, Buddhist and Shinto priests, and ‘others of no particular aliation but who by reason of their extreme nationalist sentiments would be a danger to our security as well as others who have seen Japanese military service” (Kashima, 2003, p. 68). ere was also additional wartime de- tention and relocation of some 1,500 Kibei residents—that is, Japanese-Americans living in Hawaii who had been educated in Japan, spoke primarily Japanese, or had other close personal or nan- cial interests in Japan. e result was that by far smaller numbers and shares of Hawaii’s Japanese or Japanese- American population were interned than on the mainland. Kashima (2003, pp. 86–87) summarizes:
Under this selective policy, an additional 1,217 persons eventually were shipped to WRA relocation camps from November 23, 1942 to July 1945. e largest group, numbering 1,037, arrived between November 1942 and March 14, 1943. [T]he army put all these Hawaii residents in WRA camps; most of them went to Jerome, Arkansas, or Topaz, Utah, but some ended up at Minidoka, Idaho, and Tule Lake, California. [A]dding this count of 1,217 to the 875 predominately male Issei produces a total of 2,092 persons brought to the mainland. Along with the estimated 300 persons who remained in Hawaii, the total number of Hawaii residents of Japanese ancestry were detained in permanent imprisonment facilities comes to 2,392. Compared to the prisoner population of Nikkei on the mainland, this is a very small number. e contrast in treatment between these two groups is stark and revealing.
For our purposes, the number of people interned—around 2,300 in total—represents around 2 per- cent of the overall Hawaiian Japanese or Japanese American population. By comparison, around 120,000 Japanese Americans—or around 85 percent—of people of Japanese descent living in the mainland were interned. In sum, there was never wholesale evacuation and internment of people of Japanese descent, like there was on the west coast of the United States. For these reasons, we set Hawaii aside in our analyses.
2 8.2 From Assembly Centers to Internment Camps
Over 90,000 of the 120,000 Japanese Americans in WRA custody during WWII were trans- ferred from assembly centers to internment camps throughout the fall of 1942. Notably, not all Japanese Americans who were interned transferred from assembly centers. Japanese-Americans forced to report to the Owens Valley (10,000) and Parker Dam (12,000) reception centers were not transferred to internment camps because Owens Valley was expanded to become the Manzanar internment camp and Parker Dam became Poston. Over 17,000 internees were taken to intern- ment camps directly from their homes, while 5,918 were born in internment camps, and over 4,500 were transferred to internment camps from Immigration and Naturalization Service (INS) facilities, prisons, and other institutions.14 By July 1, 1942, the WRA had completed construction on just three of the ten planned Japanese Internment camps: Manzanar, Poston, and Tule Lake. Manzanar had reached its population ca- pacity by summer 1942, but the other camps began receiving groups of internees virtually as soon as each was operational - and sometimes before they were fully constructed (WRA, 1942). e WRA’s objective was to “relocate the evacuated people as quickly as possible in order that their removal from private life and productive work would be brief” (WRA, 1943). Internees were transported to the nearest internment camps accessible by train as soon as those camps were open. In addition to speed, the WRA prioritized keeping families and communities together - the reason for this is that the Army envisioned internment camps as self contained communities where internees would work, farm, worship, run their own schools, and maintain some respon- sibility over self governance (WRA, 1943). To this end, the Army tried to relocate each assembly center’s whole population to a single nearby internment camp in one mass evacuation per as- sembly center. e WRA only deviated from this when an assembly center’s population was too large for any single internment camp to accommodate (see WRA (1943), p. 3). Internees “boarded trains at the assembly centers and travelled hundreds of miles farther
14National Japanese American Memorial Foundation. “Forced Removal and Incarceration.” https://www. njamemorial.org/internment-overview
3 inland to the partially completed relocation centers” (WRA, 1942). In interviews and oral histories, internees shared powerful memories of mass transport. “ere were guards at the end of each car and the shades were drawn,” recalls Kanji Sahara, now 82, of his removal from the Santa Anita assembly center to the internment camp in Jerome, Arkansas. Sahara and his family had reported to their local church under the initial evacuation orders and travelled to Santa Anita in a caravan of “10 or 15 buses;” the family stayed in the horse stables at Santa Anita for six months before being moved to Arkansas (Fares, 2017).
4 8.3 Internment Status (Table)
Table A.5: Internment Status and Political Engagement
Dependent variable: Political Political Political Leadership Interest Distrust Advice Approach OLS OLS OLS OLS (1) (2) (3) (4) Direct Exposure to Internment −.133∗∗∗ .043 −.019 .109∗ (.043) (.036) (.027) (.056) Family-Only Exposure to Internment −.181∗∗∗ .022 −.005 .190∗∗∗ (.046) (.036) (.027) (.057) Issei, Nisei, Nisei and Nisei and Nisei and Generations and Sansei Sansei Sansei Sansei Generation Fixed Eects XXXX Pre-Internment Location Fixed Eects XXXX Controls Age, Age, Age, Age, Gender Gender Gender Gender Observations 4,101 3,068 3,069 2,982 R2 .336 .011 .029 .038 Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
5 8.4 Formal Tests of Coecient Equality
Table A.6: Formal Tests of Coecient Equality (Direct Exposure = Family-Only Exposure)
Model Residual DF RSS df Sum of Sq. F-statistic p Political Interest 4062.00 1899.90 Political Interest 4061.00 1899.14 1.00 0.77 1.64 0.20 Political Distrust 3035.00 743.05 Political Distrust 3034.00 742.92 1.00 0.13 0.54 0.46 Political Advice 3036.00 417.96 Political Advice 3035.00 417.91 1.00 0.05 0.40 0.53 Leadership Approach 2951.00 1716.64 Leadership Approach 2950.00 1714.86 1.00 1.78 3.06 0.08
6 8.5 Internment Length (Table)
Table A.7: Internment Length and Political Engagement
Dependent variable: Political Political Leadership Interest Distrust Advice Approach OLS OLS OLS OLS (1) (2) (3) (4) Internment Length −.014 .042∗∗∗ −.034∗∗∗ .043∗∗ (.013) (.011) (.008) (.017) Issei, Nisei, Nisei and Nisei and Nisei and Generations and Sansei Sansei Sansei Sansei Generation Fixed Eects XXXX Controls Age, Gender Age, Gender Age, Gender Age, Gender Observations 2,690 1,891 1,892 1,858 R2 .360 .011 .030 .027 Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
7 8.6 Eects of Internment Experience (Table)
Table A.8: Eects of Experiencing Violence while Interned
Dependent variable:
Political Interest Political Distrust Political Advice Leadership Approach
(1) (2) (3) (4)
Violence −.064 .097∗∗∗ −.016 .084 (.059) (.022) (.038) (.051)
Interned −.011 .034 −.021 −.079∗ (.063) (.032) (.031) (.047)
Gender (Male) .135∗∗∗ −.014 .076∗∗∗ −.135∗∗∗ (.026) (.016) (.021) (.019)
Age −.001 −.001 −.004∗∗ .009∗∗∗ (.003) (.001) (.002) (.002)
Violence x Interned −.012 −.078∗∗∗ −.004 −.006 (.072) (.027) (.044) (.060)
Generations Issei, Nisei, Sansei Nisei and Sansei Nisei and Sansei Nisei and Sansei Generation Fixed Eects XXXX Pre-Internment Location Fixed Eects XXXX Observations 3,188 2,392 2,396 2,334 R2 .314 .009 .029 .027
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Note: Standard errors clustered by internment camp to which respondents are assigned.
8 Table A.9: Eects of Experiencing Strikes while Interned
Dependent variable:
Political Interest Political Distrust Political Advice Leadership Approach
(1) (2) (3) (4)
Strikes −.087 .086∗∗∗ −.039 .035 (.055) (.026) (.045) (.048)
Interned −.032 .016 −.029 −.095∗ (.056) (.037) (.035) (.050)
Gender (Male) .133∗∗∗ −.014 .076∗∗∗ −.133∗∗∗ (.026) (.017) (.021) (.019)
Age −.001 −.001 −.004∗∗ .009∗∗∗ (.003) (.001) (.002) (.002)
Strikes x Interned .038 −.010 .015 .045 (.079) (.024) (.049) (.046)
Generations Issei, Nisei, Sansei Nisei and Sansei Nisei and Sansei Nisei and Sansei Generation Fixed Eects XXXX Pre-Internment Location Fixed Eects XXXX Observations 3,188 2,392 2,396 2,334 R2 .314 .013 .030 .027
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Note: Standard errors clustered by internment camp to which respondents are assigned.
9 Table A.10: Eects of Experiencing Use of Force while Interned
Dependent variable:
Political Interest Political Distrust Political Advice Leadership Approach
(1) (2) (3) (4)
Force .061 −.018 .066 .010 (.059) (.054) (.041) (.080)
Interned −.001 .003 −.002 −.078 (.056) (.032) (.035) (.049)
Gender (Male) .133∗∗∗ −.013 .076∗∗∗ −.132∗∗∗ (.026) (.018) (.021) (.019)
Age −.001 −.001 −.004∗∗ .009∗∗∗ (.003) (.001) (.002) (.002)
Force x Interned −.067 .042 −.086∗ .006 (.108) (.047) (.047) (.071)
Generations Issei, Nisei, Sansei Nisei and Sansei Nisei and Sansei Nisei and Sansei Generation Fixed Eects XXXX Pre-Internment Location Fixed Eects XXXX Observations 3,188 2,392 2,396 2,334 R2 .313 .008 .030 .025
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Note: Standard errors clustered by internment camp to which respondents are assigned.
10 Table A.11: Eects of Militarized Camp Environment
Dependent variable:
Political Interest Political Distrust Political Advice Leadership Approach
(1) (2) (3) (4)
Militarism .116∗∗∗ −.064 .038 −.025 (.043) (.050) (.035) (.048)
Interned −.001 −.012 −.016 −.115 (.038) (.032) (.032) (.089)
Gender (Male) .135∗∗∗ −.010 .077∗∗∗ −.135∗∗∗ (.027) (.017) (.021) (.019)
Age −.001 −.001 −.004∗∗∗ .009∗∗∗ (.003) (.002) (.002) (.002)
Militarism x Interned −.033 .036 −.011 .021 (.071) (.023) (.046) (.082)
Generations Issei, Nisei, Sansei Nisei and Sansei Nisei and Sansei Nisei and Sansei Generation Fixed Eects XXXX Pre-Internment Location Fixed Eects XXXX Observations 3,135 2,357 2,360 2,299 R2 .314 .009 .030 .028
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Note: Standard errors clustered by internment camp to which respondents are assigned.
8.7 Balance Test Results
8.7.1 K-S Tests
Plots of age, gender, and pre-internment occupations for internees held at dierent camps (dis- played in Figures A.9 and A.10 and Table A.14) reveal no signicant dierences in these distribu-
11 tions across dierent internment locations. We further conduct a series of Kolmogorov-Smirnov (KS) tests of the null hypothesis that observed age and sector of employment for internees im- prisoned across dierent camps are draws from the same distribution broadly support the idea that internees are comparable across these dimensions. Note that, since our objective is to inves- tigate whether or not the groups of internees physically sent to various internment camps are in fact similar across a variety of observable characteristics, all of the analyses in this section con- dition on respondents who report having been interned themselves. Respondents who were not interned are not included in this analysis. ere are 45 possible pairs of the 10 major camps to which Japanese Americans were assigned during the internment period. For age, our KS test results reject the null hypothesis of a single parent distribution in two of 45 cases (internees in Topaz vs. Manzanar and Topaz vs. Jerome). is is with the dened Type I error rate of 0.05 for these tests. For employment sector, we treat the 3-digit designations for occupational sector as a continuous covariate and test the distributions of these sectors for Nisei respondents across camps. Here, we reject the null hypothesis of a single distribution at the 5% level in none of the 45 cases. Full results for these tests are presented in Tables A.12 and A.13. A series of t-tests for standardized dierences in proportions never reject the null hypothesis that gender balance is nearly identical across all 45 possible pairs of internment camps at the 5% level. e following gures and tables summarize balance in the distributions of age, gender, and employment sector among internees imprisoned across all 45 possible unique pairs of internment camps.
12 Table A.12: Kolmogorov-Smirnov Test Results for Age
Camp 1 Camp 2 p-value
Rivers, AZ Poston, AZ 0.53 Amache, CO Poston, AZ 0.66 Denson, AR Poston, AZ 0.43 Manzanar, CA Poston, AZ 0.26 McGehee, AR Poston, AZ 0.66 Heart Mountain, WY Poston, AZ 0.87 Hunt, ID Poston, AZ 0.63 Newell, CA Poston, AZ 0.78 Topaz, UT Poston, AZ 0.08 Amache, CO Rivers, AZ 0.52 Denson, AR Rivers, AZ 0.77 Manzanar, CA Rivers, AZ 1.00 McGehee, AR Rivers, AZ 0.74 Heart Mountain, WY Rivers, AZ 0.64 Hunt, ID Rivers, AZ 0.64 Newell, CA Rivers, AZ 0.87 Topaz, UT Rivers, AZ 0.06 Denson, AR Amache, CO 0.14 Manzanar, CA Amache, CO 0.25 McGehee, AR Amache, CO 0.72 Heart Mountain, WY Amache, CO 1.00 Hunt, ID Amache, CO 0.90 Newell, CA Amache, CO 0.53 Topaz, UT Amache, CO 0.52 Manzanar, CA Denson, AR 0.81 McGehee, AR Denson, AR 0.44 Heart Mountain, WY Denson, AR 0.17 Hunt, ID Denson, AR 0.16 Newell, CA Denson, AR 0.46 Topaz, UT Denson, AR 0.04 McGehee, AR Manzanar, CA 0.61 Heart Mountain, WY Manzanar, CA 0.37 Hunt, ID Manzanar, CA 0.31 Newell, CA Manzanar, CA 0.44 Topaz, UT Manzanar, CA 0.05 Heart Mountain, WY McGehee, AR 0.82 Hunt, ID McGehee, AR 0.68 Newell, CA McGehee, AR 0.49 Topaz, UT McGehee, AR 0.22 Hunt, ID Heart Mountain, WY 1.00 Newell, CA Heart Mountain, WY 0.56 Topaz, UT Heart Mountain, WY 0.15 Newell, CA Hunt, ID 0.92 Topaz, UT Hunt, ID 0.14 Topaz, UT Newell, CA 0.06
13 Table A.13: Kolmogorov-Smirnov Test for Employment Sector
Camp 1 Camp 2 p-value
Rivers, AZ Poston, AZ 0.09 Amache, CO Poston, AZ 0.10 Denson, AR Poston, AZ 0.24 Manzanar, CA Poston, AZ 0.22 McGehee, AR Poston, AZ 0.98 Heart Mountain, WY Poston, AZ 0.15 Hunt, ID Poston, AZ 0.10 Newell, CA Poston, AZ 0.97 Topaz, UT Poston, AZ 0.62 Amache, CO Rivers, AZ 0.93 Denson, AR Rivers, AZ 1.00 Manzanar, CA Rivers, AZ 0.68 McGehee, AR Rivers, AZ 0.46 Heart Mountain, WY Rivers, AZ 1.00 Hunt, ID Rivers, AZ 1.00 Newell, CA Rivers, AZ 0.19 Topaz, UT Rivers, AZ 0.09 Denson, AR Amache, CO 1.00 Manzanar, CA Amache, CO 1.00 McGehee, AR Amache, CO 0.68 Heart Mountain, WY Amache, CO 1.00 Hunt, ID Amache, CO 1.00 Newell, CA Amache, CO 0.12 Topaz, UT Amache, CO 0.24 Manzanar, CA Denson, AR 0.98 McGehee, AR Denson, AR 0.73 Heart Mountain, WY Denson, AR 1.00 Hunt, ID Denson, AR 1.00 Newell, CA Denson, AR 0.20 Topaz, UT Denson, AR 0.52 McGehee, AR Manzanar, CA 0.94 Heart Mountain, WY Manzanar, CA 0.97 Hunt, ID Manzanar, CA 0.96 Newell, CA Manzanar, CA 0.25 Topaz, UT Manzanar, CA 0.67 Heart Mountain, WY McGehee, AR 0.53 Hunt, ID McGehee, AR 0.66 Newell, CA McGehee, AR 0.93 Topaz, UT McGehee, AR 1.00 Hunt, ID Heart Mountain, WY 1 Newell, CA Heart Mountain, WY 0.15 Topaz, UT Heart Mountain, WY 0.11 Newell, CA Hunt, ID 0.12 Topaz, UT Hunt, ID 0.10 Topaz, UT Newell, CA 0.41
14 Table A.14: T-Tests for Dierences in Gender Proportion by Camp
Camp 1 Camp 2 Dierence in Means Standard Error T p-value Lower Bound Upper Bound
Rivers, AZ Poston, AZ 0.03 7.11 0.004 1.00 -13.99 14.00 Amache, CO Poston, AZ 0.01 7.10 0.001 1.00 -13.97 13.97 Denson, AR Poston, AZ -0.07 5.24 -0.01 0.99 -10.37 10.35 Manzanar, CA Poston, AZ -0.06 6.83 -0.01 0.99 -13.45 13.43 McGehee, AR Poston, AZ -0.01 6.18 -0.001 1.00 -12.19 12.18 Heart Mountain, WY Poston, AZ 0.03 7.35 0.003 1.00 -14.45 14.46 Hunt, ID Poston, AZ -0.01 7.58 -0.001 1.00 -14.91 14.90 Newell, CA Poston, AZ -0.02 6.95 -0.003 1.00 -13.68 13.67 Topaz, UT Poston, AZ 0.01 6.60 0.002 1.00 -12.99 12.99 Amache, CO Rivers, AZ -0.02 6.49 -0.004 1.00 -12.76 12.76 Denson, AR Rivers, AZ -0.10 4.99 -0.02 0.98 -9.88 9.84 Manzanar, CA Rivers, AZ -0.09 6.27 -0.01 0.99 -12.36 12.33 McGehee, AR Rivers, AZ -0.04 5.77 -0.01 1.00 -11.38 11.37 Heart Mountain, WY Rivers, AZ -0.004 6.67 -0.001 1.00 -13.13 13.13 Hunt, ID Rivers, AZ -0.04 6.84 -0.01 1.00 -13.47 13.46 Newell, CA Rivers, AZ -0.05 6.37 -0.01 0.99 -12.55 12.53 Topaz, UT Rivers, AZ -0.02 6.10 -0.003 1.00 -12.01 12.00 Denson, AR Amache, CO -0.08 4.98 -0.02 0.99 -9.86 9.83 Manzanar, CA Amache, CO -0.07 6.26 -0.01 0.99 -12.33 12.31 McGehee, AR Amache, CO -0.01 5.76 -0.002 1.00 -11.35 11.35 Heart Mountain, WY Amache, CO 0.02 6.66 0.003 1.00 -13.11 13.11 Hunt, ID Amache, CO -0.01 6.83 -0.002 1.00 -13.44 13.44 Newell, CA Amache, CO -0.03 6.36 -0.004 1.00 -12.52 12.51 Topaz, UT Amache, CO 0.01 6.09 0.001 1.00 -11.99 11.99 Manzanar, CA Denson, AR 0.01 4.83 0.002 1.00 -9.55 9.55 McGehee, AR Denson, AR 0.06 4.62 0.01 0.99 -9.12 9.14 Heart Mountain, WY Denson, AR 0.10 5.07 0.02 0.99 -10.00 10.04 Hunt, ID Denson, AR 0.06 5.13 0.01 0.99 -10.14 10.16 Newell, CA Denson, AR 0.05 4.91 0.01 0.99 -9.70 9.72 Topaz, UT Denson, AR 0.08 4.80 0.02 0.99 -9.48 9.51 McGehee, AR Manzanar, CA 0.06 5.59 0.01 0.99 -11.00 11.02 Heart Mountain, WY Manzanar, CA 0.09 6.44 0.01 0.99 -12.66 12.68 Hunt, ID Manzanar, CA 0.05 6.59 0.01 0.99 -12.95 12.97 Newell, CA Manzanar, CA 0.04 6.14 0.01 0.99 -12.08 12.09 Topaz, UT Manzanar, CA 0.07 5.91 0.01 0.99 -11.62 11.65 Heart Mountain, WY McGehee, AR 0.03 5.89 0.01 1.00 -11.61 11.62 Hunt, ID McGehee, AR -0.002 6.01 -0.0004 1.00 -11.84 11.84 Newell, CA McGehee, AR -0.02 5.67 -0.003 1.00 -11.18 11.18 Topaz, UT McGehee, AR 0.02 5.49 0.004 1.00 -10.81 10.82 Hunt, ID Heart Mountain, WY -0.03 7.05 -0.005 1.00 -13.88 13.87 Newell, CA Heart Mountain, WY -0.05 6.54 -0.01 0.99 -12.88 12.87 Topaz, UT Heart Mountain, WY -0.01 6.24 -0.002 1.00 -12.30 12.29 Newell, CA Hunt, ID -0.01 6.70 -0.002 1.00 -13.18 13.18 Topaz, UT Hunt, ID 0.02 6.38 0.003 1.00 -12.56 12.57 Topaz, UT Newell, CA 0.04 5.99 0.01 1.00 -11.80 11.81
15 e following gures and tables summarize balance in the distributions of age, and employ- ment sector among internees imprisoned in the largest internment camps in our data.
Figure A.9: Age Distribution for Japanese Internees by Camp
Tule Lake
Topaz
Rohwer
Poston
Minidoka
Manzanar
Jerome
Heart Mountain
Gila River
Amache
30 60 90 Age
16 Figure A.10: Distribution of Pre-Internment Occupational Sectors by Camp
Tule Lake
Topaz
Rohwer
Poston
Minidoka
Manzanar
Jerome
Heart Mountain
Gila River
Amache
0 500 1000 Employment Sector Prior to Internment
17 8.7.2 Comparing Japanese Americans Living In- and Outside of the Exclusion Zone in 1940 Using the Census
One important limitation of the JARP survey data is that it allows researchers only a narrow glimpse at respondent level characteristics assigned before respondents were interned (or not). We control for gender, age, and pre-internment locations throughout this study, but these con- trols still allow for the possibility that Japanese Americans who were interned still dier from Japanese Americans who were not along some important covariate. Because social scientists typically construe an individual’s level of political engagement as a function of socioeconomic status (Verba, Schlozman and Brady, 1995), we might be particularly worried that internees and non-internees dier along other socioeconomic covariates such as education, marital status, or specic occupation. If the population of Japanese Americans living in the Exclusion Zones on the eve of the internment period was already less educated, less wealthy or less likely to be married, skeptical researchers might argue that this population would be less politically engaged than its counterpart further inland regardless of the internment experience. In the analyses that follow, we use data from the 1940 U.S. Census to demonstrate that this is unlikely to be the case; Japanese Americans living in- and outside of the Exclusion Zone are comparable along a variety of important socioeconomic dimensions. ese analyses rely on the 1940 Census summary report entitled “Characteristics of the Nonwhite Population by Race.”15 is summary report provides distributions of age, gender, marital status, education, employment status, and employment type for Japanese Americans living in California, Oregon, Washington State (inside the Exclusion Zone), Colorado, New York, and Utah (outside the Exclusion Zone). We code Washington and Oregon as states within the Exclusion Zone because, while internment orders did not cover the entirety of both states, most Japanese Americans living in these states would have lived close to the coast and in the Exclusion Zone. e Census summarizes data for these states because each contains a Japanese-American population of at least 2,000 people. While this data does not represent the totality of Japanese Americans living in the United States in 1940,
15United States Department of Commerce. Sixteenth Census of the United States. 1940. Tables 33-38.
18 these six states represent 119,835 of the 126,947 Japanese Americans living in the United States in 1940. We present summary statistics for Japanese Americans by state in Table A.15.
Table A.15: Japanese Americans by State, 1940
State Exclusion Zone Population Prop. Male Prop. Married Prop. Under 20 Pop. Completed HS Median Yrs. Ed. Prop. Employed Prop. Professional Prop. Farmers CA 1 93, 717 56.07 49.84 25.49 22.83 8.60 55.75 2.38 14.34 CO 0 2, 734 55.05 42.23 31.49 15.93 8.10 44.57 2.84 67.92 NY 0 2, 538 70.17 50.83 17.14 21.72 12.30 63.93 6.64 0.59 OR 1 4, 071 55.78 50.08 23.63 22.48 8.50 54.91 2.54 47.91 UT 0 2, 210 57.15 44.45 29.05 17.57 8.10 47.72 2.04 25.95 WA 1 14, 565 55.56 47.49 23.08 28.49 9 56.36 2.32 12.92
ese summary statistics reveal no major dierences between Japanese Americans living in- side the Exclusion Zone and outside of the Exclusion Zone. Table A.16 displays the results of a more formal test of these dierences in the form of a set of two sided hypothesis tests for the dierences in means across a broad set of covariates that might inuence downstream political behavior. e lowest p-value corresponding to any of these tests is 0.33, which suggests that Japanese Americans living in- and outside of the Exclusion Zone are not signicantly dierent in terms of gender balance, marital status, age, education, or employment at the 95% level.
Table A.16: p Values from Two-Sided Dierence-in-Means Tests on Exclusion Zone Boundary in 1940
Variable p Prop. Male 0.40 Prop. Married 0.33 Prop. Under 20 0.72 Prop. Completed HS 0.81 Median Yrs. Ed. 0.63 Prop. Employed 0.61 Prop. Professional 0.42 Prop. Farmers 0.79
19 8.8 Assignment to Places with Particular Political or Demographic Character- istics
One alternative pathway through which internment may have aected political outcomes for internees might have involved the political or demographic characteristics of the jurisdictions that housed internment camps. In a period when wartime uncertainty and racial anti-Japanese sen- timent were at an all-time high in the mass public, skeptical readers might worry that Japanese- Americans interned in particularly white areas or areas that showed broad support for Franklin Roosevelt might have experienced more hostility. In short, the climate outside of the camps might have conveyed more of the eects we report than the climate within the camps themselves. Be- low, we provide evidence that this was unlikely to have been the case. First, qualitative accounts suggest that Japanese-Americans interned throughout the United States had extremely limited contact with the outside world. Internees would have lile information about the surrounding people living in areas with even extremely distinct political or demographic characteristics. To rule out this possibility, we included indicators for Franklin D. Roosevelt’s 1940 share of the two party vote and for the percentage white in the county of internment hosting each internment camp. ese covariates themselves are rarely signicant. e percentage white in each county of internment is only signicantly associated with level of political distrust when we consider one type of camp experience (experiencing violence). While presidential vote share in 1940 seems to consistently and signicantly inuence political communication and views on leadership ap- proach, these values are substantively small. More importantly, including these covariates does not change our substantive conclusions about the eects of experiencing violence or strikes while interned.
20 Table A.17: Eects of Experiencing Violence while Interned: Controlling for Surrounding Camp Environment
Dependent variable:
Political Interest Political Distrust Political Advice Leadership Approach
(1) (2) (3) (4)
Violence −.057 .084∗∗∗ −.022 .052 (.063) (.029) (.035) (.035)
Interned −.010 .032 −.022 −.082∗ (.063) (.033) (.032) (.047)
Gender (Male) −.001∗∗∗ .001 −.0004 .0003 (.001) (.001) (.0003) (.001)
Age −.081∗ .114 .033 .221∗∗∗ (.045) (.113) (.022) (.066)
Violence x Interned .135∗∗∗ −.015 .075∗∗∗ −.137∗∗∗ (.026) (.016) (.020) (.020)
FDR 1940 Vote Share −.001 −.0005 −.004∗∗ .009∗∗∗ (.003) (.002) (.002) (.002)
Pct. White 1940 −.012 −.077∗∗∗ −.004 −.007 (.070) (.027) (.044) (.063)
Generations Issei, Nisei, Sansei Nisei and Sansei Nisei and Sansei Nisei and Sansei Generation Fixed Eects XXXX Pre-Internment Location Fixed Eects XXXX Observations 3,188 2,392 2,396 2,334 R2 .314 .011 .030 .030
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Note: Standard errors clustered by internment camp to which respondents are assigned.
21 Table A.18: Eects of Experiencing Strikes while Interned: Controlling for Surrounding Camp Environment
Dependent variable:
Political Interest Political Distrust Political Advice Leadership Approach
(1) (2) (3) (4)
Strikes −.072 .079∗∗∗ −.043 .020 (.060) (.028) (.040) (.037)
Interned −.030 .014 −.030 −.103∗∗ (.055) (.037) (.035) (.051)
Gender (Male) −.001 .0001 .00004 −.0004 (.001) (.001) (.001) (.001)
Age −.097∗ .070 .041 .210∗∗∗ (.058) (.061) (.042) (.047)
Strikes x Interned .134∗∗∗ −.015 .075∗∗∗ −.136∗∗∗ (.026) (.017) (.020) (.020)
FDR 1940 Vote Share −.001 −.001 −.004∗∗ .009∗∗∗ (.002) (.001) (.002) (.002)
Pct. White 1940 .036 −.007 .017 .053 (.076) (.024) (.048) (.047)
Generations Issei, Nisei, Sansei Nisei and Sansei Nisei and Sansei Nisei and Sansei Generation Fixed Eects XXXX Pre-Internment Location Fixed Eects XXXX Observations 3,188 2,392 2,396 2,334 R2 .314 .014 .030 .030
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Note: Standard errors clustered by internment camp to which respondents are assigned.
22 Table A.19: Eects of Experiencing Use of Force while Interned: Controlling for Surrounding Camp Environment
Dependent variable:
Political Interest Political Distrust Political Advice Leadership Approach
(1) (2) (3) (4)
Force .164∗ −.089∗∗ .068 −.108∗∗ (.093) (.035) (.045) (.047)
Interned .002 .0002 −.002 −.084∗ (.056) (.032) (.035) (.050)
Gender (Male) −.004∗∗∗ .002∗∗∗ −.0005 .002∗∗ (.001) (.001) (.001) (.001)
Age −.344∗∗∗ .245∗∗ .007 .463∗∗∗ (.068) (.098) (.065) (.115)
Force x Interned .135∗∗∗ −.016 .076∗∗∗ −.138∗∗∗ (.026) (.017) (.021) (.020)
FDR 1940 Vote Share −.001 −.0004 −.004∗∗ .010∗∗∗ (.003) (.002) (.002) (.002)
Pct. White 1940 −.061 .038 −.087∗ −.004 (.113) (.045) (.047) (.069)
Generations Issei, Nisei, Sansei Nisei and Sansei Nisei and Sansei Nisei and Sansei Generation Fixed Eects XXXX Pre-Internment Location Fixed Eects XXXX Observations 3,188 2,392 2,396 2,334 R2 .315 .011 .031 .031
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Note: Standard errors clustered by internment camp to which respondents are assigned.
23 Table A.20: Eects of Militarized Camp Environment: Controlling for Surrounding Camp Envi- ronment
Dependent variable:
Political Interest Political Distrust Political Advice Leadership Approach
(1) (2) (3) (4)
Militarism .118∗∗ −.066 .038 −.027 (.053) (.042) (.034) (.032)
Interned −.004 −.009 −.016 −.113 (.038) (.029) (.033) (.091)
Gender (Male) −.002∗∗∗ .001 −.001∗∗∗ .0001 (.0002) (.001) (.0001) (.001)
Age −.174∗∗∗ .148 −.006 .248∗∗∗ (.030) (.095) (.021) (.039)
Militarism x Interned .136∗∗∗ −.012 .077∗∗∗ −.139∗∗∗ (.027) (.017) (.021) (.020)
FDR 1940 Vote Share −.001 −.001 −.004∗∗ .010∗∗∗ (.003) (.002) (.002) (.002)
Pct. White 1940 −.023 .028 −.011 .008 (.069) (.021) (.046) (.083)
Generations Issei, Nisei, Sansei Nisei and Sansei Nisei and Sansei Nisei and Sansei Generation Fixed Eects XXXX Pre-Internment Location Fixed Eects XXXX Observations 3,135 2,357 2,360 2,299 R2 .315 .011 .031 .032
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Note: Standard errors clustered by internment camp to which respondents are assigned.
24 8.9 Recording of Pre-Internment Location
One important limitation for the JARP data is the fact that pre-internment location, in most cases outside of California, is recorded at the state level.16 While it is plausible to assume that peo- ple living in a single city or county area would have had similar probabilities of camp assignment, it is less clear that this assumption might hold for everyone living in a given state. To address this issue, we examine individual internee records provided by the WRA. We use this data to simulate more specic pre-internment locations for JARP respondents. We took the following approach. First, we used the WRA data to calculate the empirical probabilities of having lived in specic areas within a state. For instance, according to the WRA records 6 Japanese Americans interned aer 1941 had last-known addresses in Colorado before they were relocated: 4 in Rocky Falls, 1 in Fort Upton, and 1 in Greeley. For JARP respondents living in a given state, we drew a specic location from a multinomial distribution using these empirical probabilities at the family level. In our example, every member of a single family unit living in the state of Colorado would have a 67% probability of being assigned to Rocky Falls and 16% probabilities of being assigned to Fort Upton or Greely, respectively. Not all family members in the JARP shared a pre-internment location, but we assumed that family members who did share a pre-internment code in the JARP survey would share a specic location. is means that we assumed all members of a given family living in Washington state in the JARP data might be assigned to Seale if this was the specic location drawn for their family ID (but members of the same family living in California would all get a dierent, single assigned location in California)17. We assigned every respondent in the JARP who was interned to a more specic pre-internment location using this procedure and re-estimated the eects of internment experience using the
16For California, JARP provides separate designations for people who lived in the state’s largest Metropolitan Sta- tistical Areas (“MSA”) from people who lived in “rural clusters” throughout the state. ese designations provide lile additional information relative to the statewide code because they group jurisdictions across geographic areas. San Francisco and Los Angeles, for instance, are coded using the same value because both are large MSAs 17Specic location assignments for California respondents are generated na¨ıvely, meaning that we use the raw empirical probabilities of living in any WRA-designated sub-area of California instead of mapping between the JARP’s regional California codes and the WRA’s. e reason for this is that the lists of locations within California do not match exactly between the WRA and the JARP, and it is unclear how to assign specic locations to the ”other California” category listed in the JARP
25 Figure A.11: Eect of Witnessing Strikes While Interned Using WRA Location Data
Effect of Witnessing Strikes While Interned
● Political Interest ●
● Political Distrust ●
Group
● Direct Exposure ● Family−Only Exposure Outcome
● Political Advice ●
● Leadership Approach ●
−0.1 0.0 0.1 Coefficient Estimate same approach described in Section 6. We repeated this procedure 1,000 times to get a distribution of eect sizes under a variety of assumed pre-internment locations for JARP survey respondents. Figures A.11 through A.14 display the results of these simulations. e means of each plot- ted distribution can be interpreted as the average eect of witnessing violence, force, strikes, or heavily militarized camp conditions across 10,000 possible distributions of internees in the JARP across specic pre-internment locations, where distributions across locations are based on the real places in which Japanese Americans lived on the eve of internment as recorded by the WRA. We compare the coecients and standard errors (for these results, standard errors are the standard deviations of the distribution of coecients generated for each of 10,000 simulations) for these simulations to our original results. e simulation produces very similar point estimates for the eects of various camp characteristics even under dierent assumed pre-internment locations, which provides support for plausible exogeneity of camp assignment.
26 Figure A.12: Eect of Witnessing Violence While Interned Using WRA Location Data
Effect of Witnessing Violence While Interned
● Political Interest ●
● Political Distrust ●
Group
● Direct Exposure ● Family−Only Exposure Outcome
● Political Advice ●
● Leadership Approach ●
−0.1 0.0 0.1 0.2 0.3 Coefficient Estimate
Figure A.13: Eect of Witnessing Use of Force While Interned Using WRA Location Data
Effect of Witnessing Force While Interned
● Political Interest ●
● Political Distrust ●
Group
● Direct Exposure ● Family−Only Exposure Outcome
● Political Advice ●
● Leadership Approach ●
−0.1 0.0 0.1 Coefficient Estimate
27 Figure A.14: Eect of Militarized Camp Environment using WRA Location Data
Effect of Militarized Camp Environment
● Political Interest ●
● Political Distrust ●
Group
● Direct Exposure ● Family−Only Exposure Outcome
● Political Advice ●
● Leadership Approach ●
−0.2 −0.1 0.0 0.1 Coefficient Estimate
9 Additional Robustness Checks
9.1 Military Service
Nisei and Sansei respondents were asked whether they or their spouses had ever served in the U.S. armed forces. ere is some missingness in this variable: 179 of 2,304 Nisei have no recorded response for military service, as do 347 of 802 Sansei. Our results do not suggest that specic features of internment camps produced meaningful dierences in Japanese-American military service. Similarly, regressing this indicator for military service on measures of internment expo- sure suggests internment did not produce dierences in rates of military service among interned and non-interned Japanese Americans. Moreover, the coecient estimates for each group move in opposite directions, which cannot explain the directionally consistent estimates among both groups that we observe throughout the paper. is provides additional evidence for the fact that, prior to internment, Japanese Americans who were interned were similar to Japanese Americans who were not. One potential threat to inference might reside in the possibility that a latent form of patriotism may have both led some Japanese Americans to voluntarily relocate away from the
28 Table A.21: Internment Status and Post-Treatment Military Service
Dependent variable: Military Service OLS Direct Exposure to Internment −.042 (.028) Family-Only Exposure to Internment .051 (.034) Baseline .613∗∗∗ (.025) Observations 2,571 R2 .005 Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 coasts (thereby avoiding internment) and to serve in the military later. is does not appear to be the case in our data.
9.2 Income
Another possible threat to inference is the possibility that Japanese-Americans who were not interned had higher (or lower) earning potential than Japanese Americans who were. If so, that potential is likely to have translated into higher family income aer internment. In Table A.22, we show that there are dierences in income among those who personally and indirectly expe- rienced internment, and these dierences are statistically signicant at the .05 level. However, these estimates run in the opposite direction, and thus, cannot explain the directionally consis- tent estimates that we observe for both groups in our models. Moreover, our analysis using 1940 Census data – which examined balance on pre-treatment socioeconomic characteristics and is presented in Table A.15 – suggests that those who lived within the exclusion zone and outside of it were comparable in terms of socioeconomic characteristics.
29 Table A.22: Internment Status and Post-Treatment Income
Dependent variable: Post-Treatment Income OLS Direct Exposure to Internment .605∗∗∗ (.114) Family-Only Exposure to Internment −1.136∗∗∗ (.139) Baseline 4.412∗∗∗ (.104) Observations 3,891 R2 .068 Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
9.3 Returning Home
Our data makes it dicult to identify how many people returned home aer internment. e WRA data does not provide post-camp locations for internees. While the JARP does contain specic information on the places Japanese-American internees returned to aer WWII, their pre-camp locations are recorded with less precision. Accordingly, all we can say is whether or not Issei (the only generation for whom post-camp location is recorded) returned home to the same state and sub-state area aer internment. If the 849 Issei for whom post-camp location is recorded, 425, or just over 50%, returned to the same area within their state.
9.4 Selection into Internment Experience
e U.S. Army’s records detailing the procedure by which Japanese-Americans were assigned and transferred to internment camps suggests that internees were not assigned to camp desti- nations based on their social, economic, or political characteristics. Still, one potential source of concern might be how responses to internment and interrogation among internees ultimately af- fected their experience once they were assigned to internment camps. For instance, readers might
30 worry that internees who resisted internment might have been assigned to more punitive or more militarized facilities. One prominent example of this includes the WRA’s loyalty questionnaire - launched in the winter of 1943. Men over the age of 17 who were still incarcerated in WRA camps were asked to ll out a questionnaire that asked them whether they would comply with selective service if they were draed into the U.S. military and whether they could pledge their unquali- ed allegiance to the United States. Approximately 12,000 of 78,000 respondents to this survey answered “no” to both questions, and were sequestered at the Tule Lake camp for “disloyalty” as a result. To address possible selection issues like this, we replicated our analysis with a restriction just to respondents under the age of 17. We do this because respondents under 17 at the time of internment (we set this to 1945, the nal year of WWII) could not have taken the survey or been in the group of respondents who expressed dissatisfaction at the government’s action in this way. Similarly, younger respondents would have been unlikely sources of resistance or punishment for it. Figures A.15 through A.18 display results for this population, numbering 1,173 to 1,192 respondents. Variation in observations is a result of missingness. ese results are statistically and substantively consistent with the internment experience eects we report in the main body of this paper.
31 Figure A.15: Eect of Witnessing Strikes While Interned for Internees Younger than 17 in 1945
Effect of Witnessing Strikes While Interned
● Political Interest ●
● Political Distrust ●
Group
● Direct Exposure ● Family−Only Exposure Outcome
● Political Advice ●
● Leadership Approach ●
−0.2 −0.1 0.0 0.1 0.2 Coefficient Estimate
Figure A.16: Eect of Witnessing Violence While Interned for Internees Younger than 17 in 1945
Effect of Witnessing Violence While Interned
● Political Interest ●
● Political Distrust ●
Group
● Direct Exposure ● Family−Only Exposure Outcome
● Political Advice ●
● Leadership Approach ●
−0.2 −0.1 0.0 0.1 0.2 Coefficient Estimate
32 Figure A.17: Eect of Witnessing Use of Force While Interned for Internees Younger than 17 in 1945
Effect of Witnessing Use of Military Force While Interned
● Political Interest ●
● Political Distrust ●
Group
● Direct Exposure ● Family−Only Exposure Outcome
● Political Advice ●
● Leadership Approach ●
−0.1 0.0 0.1 0.2 Coefficient Estimate
Figure A.18: Eect of Militarized Camp Environment for Internees Younger than 17 in 1945
Effect of Being Interned in Highly Militarized Camp
● Political Interest ●
● Political Distrust ●
Group
● Direct Exposure ● Family−Only Exposure Outcome
● Political Advice ●
● Leadership Approach ●
−0.2 0.0 0.2 Coefficient Estimate
33